mradermacher/Mira-v1.12.1-27B-i1-GGUF
27B
•
Updated
•
1.14k
Datasets:
identifier
stringlengths 0
11.3k
| collection
stringclasses 118
values | open_type
stringclasses 6
values | curator
stringclasses 15
values | license
stringclasses 635
values | date
int64 1
2.1k
⌀ | title
stringlengths 0
1.59k
⌀ | creator
stringlengths 0
3.5k
⌀ | language
stringclasses 367
values | language_type
stringclasses 5
values | word_count
int64 1
650k
| token_count
int64 0
1M
| text
stringlengths 2
4.21M
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
2ca42ebc496ce529f1f0529356b1cdef
|
French Open Data
|
Open Government
|
Pleias
|
Various open data
| null |
cour-d'appel_n°2102441_08_02_2024.pdf
|
courdecassation.fr
|
French
|
Written
| 3,540
| 5,962
|
8 février 2024
Cour d'appel d'Orléans
RG n° 21/02441
Chambre Commerciale
Texte de la décision
Entête
COUR D'APPEL D'ORLÉANS
CHAMBRE COMMERCIALE, ÉCONOMIQUE ET FINANCIÈRE
GROSSES + EXPÉDITIONS : le 08/02/2024
la SAS DUVIVIER & ASSOCIES
la SELARL STRATEM AVOCATS
ARRÊT du : 08 FEVRIER 2024
N° : 36 - 23
N° RG 21/02441
N° Portalis DBVN-V-B7F-GN43
DÉCISION ENTREPRISE : Jugement du Juge des contentieux de la protection de TOURS en date du 16 Juillet 2021
Page 1 / 13
8 février 2024
PARTIES EN CAUSE
APPELANTE :- Timbre fiscal dématérialisé N°: 1265265975773543
S.A.S. SOGEFINANCEMENT
[Adresse 4]
[Localité 5]
Ayant pour avocat Me Louise BOIDIN membre de la SAS DUVIVIER & ASSOCIES, avocat au barreau de TOURS
D'UNE PART
INTIMÉ : - Timbre fiscal dématérialisé N°: 1265275448550170
Monsieur [F] [Y]
né le [Date naissance 1] 1982 à [Localité 6]
[Adresse 3]
[Localité 2]
Ayant pour avocat Me Marc ALEXANDRE de la SELARL STRATEM AVOCATS, avocat au barreau de TOURS
Page 2 / 13
8 février 2024
D'AUTRE PART
DÉCLARATION D'APPEL en date du : 15 Septembre 2021
ORDONNANCE DE CLÔTURE du : 16 novembre 2023
COMPOSITION DE LA COUR
Lors des débats, affaire plaidée sans opposition des avocats à l'audience publique du JEUDI 07 DECEMBRE 2023, à 9
heures 30, devant Madame Fanny CHENOT, Conseiller Rapporteur, par application de l'article 805 du code de procédure
civile.
Lors du délibéré :
Madame Carole CHEGARAY, Président de la chambre commerciale à la Cour d'Appel d'ORLEANS,
Madame Fanny CHENOT, Conseiller,
Monsieur Damien DESFORGES, Conseiller,
Greffier :
Madame Marie-Claude DONNAT , Greffier lors des débats et du prononcé.
ARRÊT :
Page 3 / 13
8 février 2024
Prononcé publiquement par arrêt contradictoire le JEUDI 08 FEVRIER 2024 par mise à la disposition des parties au Greffe
de la Cour, les parties en ayant été préalablement avisées dans les conditions prévues au deuxième alinéa de l'article 450
du code de procédure civile.
Exposé du litige
EXPOSE DU LITIGE :
Exposant avoir consenti à M. [F] [Y], le 2 février 2018 par voie électronique, un prêt personnel d'un montant de 30 000
euros, puis avoir vainement mis en demeure l'emprunteur de lui régler les échéances restées impayées par courrier
recommandé du 4 juillet 2019, la société Sogefinancement a provoqué la déchéance du terme de son concours le 29
août 2019, mis en demeure M. [Y] de lui payer la somme de 30'840,19'euros par courrier recommandé du même jour
réceptionné le 3 septembre suivant, puis a fait assigner M. [Y] en paiement devant le juge des contentieux de la
protection du tribunal judiciaire de Tours par acte du 31 juillet 2020.
Par jugement réputé contradictoire du 16 juillet 2021, en retenant que la société Sogefinancement ne rapportait pas la
preuve du consentement de M. [Y] à l'offre de prêt litigieuse, faute d'apporter la preuve de la signature électronique de
ce dernier et de pouvoir se prévaloir de la présomption de fiabilité définie au décret n° 2017-1416 du 28 septembre 2017,
le tribunal a':
- débouté la société Sogefinancement de l'ensemble de ses demandes dirigées contre M. [F] [Y] au titre de l'offre de prêt
personnel émise le 2 février 2018 pour un montant de 30 000 euros,
- rappelé que la présente décision est exécutoire de droit à titre provisoire,
- rappelé que le présent jugement sera non avenu s'il n'est pas notifié dans les six mois de sa date,
- condamné la société Sogefinancement aux entiers dépens.
La société Sogefinancement a relevé appel de cette décision par déclaration du 15 septembre 2021, en critiquant
Page 4 / 13
8 février 2024
expressément toutes ses dispositions lui faisant grief.
Dans ses dernières conclusions notifiées le 25 mai 2023 par voie électronique, la société Sogefinancement demande à la
cour, au visa de l'article 1134 du code civil (ancienne rédaction) et des articles L. 311 et suivants du code de la
consommation, de':
- infirmer le jugement rendu le 16 juillet 2021 par le juge du contentieux de la protection près le tribunal judiciaire de
Tours en ce qu'il a statué comme suit':
* débouté la société Sogefinancement de l'ensemble de ses demandes dirigées contre M. [F] [Y] au titre de l'offre de prêt
personnel émise le 2 février 2018 pour un montant de 30 000 euros,
* condamné la société Sogefinancement aux entiers dépens,
Statuant à nouveau,
- constater l'aveu judiciaire irrévocable de M. [F] [Y] qu'il a souscrit le crédit électronique en date du 2 février 2018,
En conséquence,
- condamner M. [F] [Y] à payer à la SAS Sogefinancement la somme de 28'577,43 euros outre les intérêts au taux
conventionnel de 5,35 % à compter de la mise en demeure du 4 juillet 2019,
- condamner M. [F] [Y] à payer à la SAS Sogefinancement la somme de 2'230,83 euros au titre de la clause pénale avec
intérêts de droit à compter du jugement à intervenir,
Subsidiairement,
- condamner M. [F] [Y] à payer à la SAS Sogefinancement la somme de 30'000 euros représentant le capital emprunté
outre les intérêts au taux légal à compter de la décision à intervenir,
En tout état de cause,
- débouter M. [F] [Y] de toutes demandes plus amples ou contraires,
- condamner M. [F] [Y] à payer à la SAS Sogefinancement la somme de 2'000 euros sur le fondement de l'article 700 du
code de procédure civile,
- condamner M. [F] [Y] en tous les dépens de première instance et d'appel.
Page 5 / 13
8 février 2024
Dans ses dernières conclusions notifiées le 19 juillet 2022 par voie électronique, M. [Y] demande à la cour, au visa des
articles 1101 et suivants du code civil, du règlement (UE) n°910/2014 du parlement européen et du conseil du 23 juillet
2014 sur l'identification électronique et les services de confiance pour les transactions électroniques au sein du marché
intérieur abrogeant la directive 1999/93/CE, du décret n°2017-1416 du 28 septembre 2017, des articles 1367 du code civil
et L. 341-2 du code de la consommation, de':
A titre principal':
- débouter la société Sogefinanement de toutes ses demandes,
- confirmer intégralement le jugement rendu par le juge des contentieux de la protection de Tours en date du 16 juillet
2021,
- débouter la société Sogefinancement de sa demande tendant à voir constater un prétendument aveu irrévocable de M.
[Y] qu'il a souscrit le crédit électronique en date du 2 février 2018,
A titre très subsidiaire':
- ordonner la déchéance du droit aux intérêts de la société Sogefinancement sur le contrat de crédit litigieux,
- débouter la société Sogefinancement de sa demande de paiement des intérêts, et déchoir Sogefinancement de tout
droit aux intérêts,
- condamner la société Sogefinancement à payer à M. [Y] une indemnité de 30'000 euros à titre de dommages et intérêts,
- débouter la société Sogefinancement de sa demande de paiement d'une indemnité «'contractuelle'» sur le contrat
litigieux,
- réduire l'indemnité contractuelle due par M. [Y] à un euro,
En tout état de cause':
- condamner la société Sogefinancement à payer à M. [Y] la somme de 3'500 euros au titre de l'article 700 du code de
procédure civile,
- condamner la société Sogefinancement aux entiers dépens de première instance et d'appel.
Pour un plus ample exposé des faits et des moyens des parties, il convient de se reporter à leurs dernières conclusions
récapitulatives.
L'instruction a été clôturée par ordonnance du 16 novembre 2023, pour l'affaire être plaidée le 7 décembre suivant et
mise en délibéré à ce jour.
Motivation
Page 6 / 13
8 février 2024
SUR CE, LA COUR :
Sur la demande en paiement de la société Sogefinancement :
Aux termes de l'article 1367 du code civil, la signature nécessaire à la perfection d'un acte juridique identifie son auteur.
Elle manifeste son consentement aux obligations qui découlent de cet acte. Quand elle est apposée par un officier public,
elle confère l'authenticité à l'acte.
Lorsqu'elle est électronique, elle consiste en l'usage d'un procédé fiable d'identification garantissant son lien avec l'acte
auquel elle s'attache. La fiabilité de ce procédé est présumée, jusqu'à preuve contraire, lorsque la signature électronique
est créée, l'identité du signataire assurée et l'intégralité de l'acte garantie, dans des conditions fixées par décret en
Conseil d'État.
Le décret n° 2017-1416 du 28 septembre 2017 relatif à la signature électronique, pris pour l'application de l'article 1367
du code civil, prévoit à son article 1er que la fiabilité d'un procédé de signature électronique est présumée jusqu'à preuve
contraire, lorsque ce procédé met en 'uvre une signature électronique "qualifiée".
Est une signature qualifiée, ainsi qu'il est précisé au second alinéa de cet article, une signature électronique avancée
conforme à l'article 26 du règlement UE n° 910/2014 du 23 juillet 2014 sur l'identification électronique et les services de
confiance pour les transactions électroniques au sein du marché [dit Règlement eIDAS] et créée à l'aide d'un dispositif de
création de signature électronique qualifié répondant aux exigences de l'article 29 dudit règlement, qui repose sur un
certificat qualifié de signature électronique répondant aux exigences de l'article 28 de ce règlement.
La présomption de fiabilité de la signature électronique, comme toute présomption, déplace l'objet de la preuve, mais ne
la supprime pas'; l'appelante n'est par conséquent pas dispensée de cette preuve.
Pour bénéficier de la présomption dont elle se prévaut, la société Sogefinancement doit rapporter la preuve de
Page 7 / 13
8 février 2024
l'existence de la signature électronique elle-même, et la preuve de sa qualification, qui passe par celle d'un dispositif de
création qualifié conformément à la définition réglementaire de la signature électronique qualifiée. Seule cette double
preuve lui permet de bénéficier de la présomption de fiabilité de la signature électronique portant sur l'intégralité de
l'acte et l'identité du signataire.
En cause d'appel, la société Sogefinancement produit, en sus de la capture d'écran du seul résumé du chemin de preuve
qui avait été communiqué en première instance (pièce 5), une attestation de signature électronique établie par la société
Idemia, prestataire de services confiance inscrit sur la liste de l'ANSSI, tiers certificateur agréé (pièce 13), outre un rapport
de certification par l'ANSSI du produit «'Dictao trust Platform'» utilisé par son prestataire de services de signature
électronique.
Alors même qu'elle reproche au premier juge d'avoir retenu que la signature électronique dont elle se prévaut ne
pouvait être considérée comme une signature «'qualifiée'» au sens de l'article 28 précité, l'appelante ne produit aucun
certificat de signature qualifié et ne justifie même pas que la signature dont elle se prévaut réponde aux exigences d'une
signature électronique seulement «'avancée'» au sens de l'article 26 du règlement eIDas.
En ne produisant pas le chemin de preuve, mais seulement, en pièce 5, deux captures d'écran de ce chemin de preuve,
sans les détails qu'il lui était loisible de télécharger, la société Sogefinancement n'établit d'aucune manière que la
signature électronique dont elle se prévaut soit liée de manière univoque à M. [Y] et permette de l'identifier.
Rien, en effet, dans les pièces communiquées, notamment dans «'l'attestation de signature électronique'» et la
«'chronologie de la transaction'», ne permet notamment de savoir comment M. [Y] se serait identifié, ni à partir de quel
appareil.
C'est à raison, dans ces circonstances, que le premier juge a retenu que la société Sogefinancement ne pouvait se
prévaloir de la présomption de fiabilité établie au seul bénéfice de la signature électronique qualifiée.
L'établissement d'une présomption de fiabilité au bénéfice de la signature qualifiée ne signifie pas que la signature
électronique non qualifiée est dépourvue de force probante. Elle constitue un moyen de preuve admissible selon l'article
1367 du code civil, mais, à défaut d'être qualifiée, il appartient à celui qui s'en prévaut d'établir sa force probante en
établissant, conformément à l'article 1367, qu'elle résulte de l'usage d'un procédé fiable d'identification garantissant son
lien avec l'acte auquel elle s'attache, c'est-à-dire de démontrer qu'elle est imputable à celui que l'on désigne comme
auteur, et qu'elle est bien attachée au document concerné.
Page 8 / 13
8 février 2024
En l'espèce, la société Sogefinancement ne peut sérieusement reprocher au premier juge d'avoir considéré que l'identité
du signataire n'était pas garantie, alors qu'en l'absence de production du fichier de preuve, elle n'offre, encore à hauteur
d'appel, aucun justificatif ni même aucun descriptif des vérifications concrètement effectuées par le prestataire de
services de confiance pour s'assurer de l'identité du signataire.
Même en l'absence de signature électronique probante, il reste que la preuve du prêt peut être rapportée selon le droit
commun.
Aux termes de l'article 1359 du code civil, l'acte juridique portant sur une somme ou une valeur excédant un montant
fixé par décret [à 1'500 euros] doit être prouvé par écrit sous signature privée ou authentique.
Selon l'article 1361 du même, il peut être suppléé à l'écrit par l'aveu judiciaire, le serment décisoire ou un
commencement de preuve par écrit corroboré par un autre moyen de preuve.
L'article 1383-2 énonce que l'aveu judiciaire est la déclaration que fait en justice la partie ou son représentant
spécialement mandaté.
S'il est exact que, en première page de ses premières écritures, le conseil de M. [Y] a écrit que «'M. [F] [Y] a souscrit
auprès de la société Sogefinancement un prêt de 30'000 euros...'», il est tout aussi certain que, dans ces mêmes
écritures, M. [Y] sollicitait la confirmation du jugement entrepris en faisant valoir que le premier juge avait justement
retenu que la société Sogefinancement n'établissait pas que le contrat litigieux avait reçu son consentement et avait été
effectivement signé par lui.
Dès lors que l'aveu exige de la part de son auteur une manifestation non équivoque de sa volonté de reconnaître pour
vrai un fait de nature à produire contre lui des conséquences juridiques, la première phrase des premières conclusions
de M. [Y] ne peut être tenue pour constitutive d'un aveu alors qu'elle résulte manifestement d'une simple erreur de
plume, qui a été corrigée dans les dernières écritures de l'intimé dans lesquelles la formule «'a souscrit'» est devenue
«'aurait souscrit'».
L'article 1362 précise que constitue un commencement de preuve par écrit tout écrit qui, émanant de celui qui conteste
un acte ou de celui qu'il représente, rend vraisemblable ce qui est allégué.
En l'espèce M. [Y] ne conteste pas avoir signé manuscritement, le 19 septembre 2018, l'avenant de réaménagement de
crédit que la société Sogefinancement produit en pièce 8.
Page 9 / 13
8 février 2024
Sur cette convention qui comporte une seule page et sur laquelle il est indiqué qu'elle a été établie en deux exemplaires,
il est rappelé, en préambule des modalités du réaménagement convenu, dans un article 1 intitué «'exposé'», ce qui suit':
«'Aux termes d'une offre préalable de crédit acceptée par l'emprunteur en date du 02.02.2018, le prêteur a consenti à
l'emprunteur, un crédit de 30'000 euros en principal aux conditions figurant dans cet acte. TEG de 5,34'%, suivant l'article
R. 341-1 du code de la consommation. Le crédit se trouvant en situation d'impayés, les parties ont convenu de procéder
aux modifications détaillées au paragraphe 2 ci-dessous, aux fins de régularisation de la situation des emprunteurs et ce,
en application des dispositions de l'article R. 312-35 alinéa 2 du code de la consommation'».
Dès lors que cet acte qui émane de M. [Y] rend vraisemblable que celui-ci ait consenti au contrat de crédit litigieux, et que
M. [Y] ne conteste, ni avoir reçu la somme de 30'000 euros, ni avoir partiellement exécuté le contrat en remboursant un
certain nombre de mensualités, ce qui ressort de l'acte de réaménagement sur lequel il est précisé qu'au 19 septembre
2018, sur le capital prêté de 30'000 euros, les sommes restant dues, intérêts impayés compris, s'élevaient à 29'079,92
euros, la société Sogefinancement rapporte la preuve de ce qu'elle a prêté à M. [Y], le 2 février 2018, la somme de 30'000
euros.
S'il est effectivement stipulé à l'avenant que celui-ci «'ne porte pas novation au contrat de crédit sus-référencé avec
lequel il forme un tout indivisible'», et qu'il «'n'annule et ne remplace que les stipulations qui lui sont contraires'», de
sorte que le premier juge a pu retenir que «'cet avenant n'est pas suffisant pour établir la signature par M. [Y] du contrat
de prêt principal'», c'est de manière inexacte en revanche qu'il en a déduit que la société Sogefinancement devait être
déboutée de toutes ses prétentions, alors que cette dernière rapporte la preuve de ce qu'elle a prêté à M. [Y] une somme
de 30'000 euros.
Faute pour la société Sogefinancement d'établir qu'elle a accordé ce prêt de 30'000 euros à M. [Y] en se conformant aux
prescriptions d'ordre public du code de la consommation, c'est-à-dire en soumettant à l'emprunteur une offre de prêt
conforme aux articles L. 312-1 et suivants de ce code, en lui fournissant les informations idoines et en procédant, en
préalable à l'octroi de ce prêt, à une évaluation de la solvabilité de l'emprunteur, la société Sogefinancement ne peut en
revanche prétendre qu'au seul remboursement du capital prêté, déduction faite des règlements effectués par M. [Y].
Dès lors, au vu de l'historique du prêt qu'il ne conteste pas, M. [Y] sera condamné à payer à la société Sogefinancement,
déduction faite de ses règlements qui s'élèvent à la somme totale de 6'131,55'euros, la somme de 23'868,45'euros,
majorée des intérêts au taux légal à compter du 3 septembre 2019, date de la mise en demeure prévue à l'article 1231-6
du code civil.
Page 10 / 13
8 février 2024
Sur la demande reconventionnelle en dommages et intérêts de M. [Y] :
En application de l'article 1231 du code civil, le dispensateur de crédit est tenu d'un devoir de mise en garde envers
l'emprunteur non averti, ou lorsqu'il a sur ses revenus, son patrimoine et ses facultés de remboursement
raisonnablement prévisibles, en l'état du succès escompté de l'opération financée, des informations que lui-même
ignorait.
La responsabilité du prêteur peut donc être engagée pour manquement à ce devoir à raison de l'inadaptation du prêt
aux capacités financières de l'emprunteur ou du risque d'endettement excessif né de l'octroi du prêt.
L'obligation de mise en garde à laquelle peut être tenu un établissement de crédit à l'égard d'un emprunteur non averti
avant de lui consentir un prêt ne porte que sur l'inadaptation de celui-ci aux capacités financières de l'emprunteur ou sur
le risque d'endettement qui résulte de son octroi, et s'apprécie à la date de l'engagement.
Il s'ensuit que le prêteur n'est tenu d'aucun devoir de mise en garde si la charge de remboursement du prêt n'excède pas
les facultés contributives de son client ou si ce dernier est un emprunteur averti.
S'il appartient à l'établissement de crédit, conformément au deuxième alinéa de l'article 1353 du code civil, de prouver
qu'il a rempli son devoir de mise en garde, encore faut-il que l'emprunteur établisse, au préalable, qu'à l'époque de la
souscription du prêt litigieux, sa situation financière justifiait l'accomplissement d'un tel devoir (v. par ex. com. 11 avril
2012, n° 11-14.507'; civ. 1, 19 décembre 2013, n° 12-20.606).
En l'espèce M. [Y], qui dénie sa signature électronique, ne peut pas sérieusement reprocher à la société Sogefinancement
d'avoir établi une fiche de dialogue incomplète sur la foi de renseignements transmis par voie électronique.
Dès lors qu'il ne fournit pas le moindre justificatif de ses ressources et charges à l'époque de l'engagement litigieux,
hormis une facture de téléphonie et un contrat de location avec option d'achat portant sur un véhicule Land Rover d'une
valeur de plus de 45'500 euros qui tend plutôt à démontrer qu'il se trouvait dans une situation financière confortable ou
se présentait comme tel, M. [Y], qui ne démontre pas que le prêt litigieux faisait naître un risque d'endettement excessif
ou était inadapté à sa situation financière, sera débouté de sa demande reconventionnelle en dommages et intérêts.
Page 11 / 13
8 février 2024
Sur les demandes accessoires :
M. [Y], qui succombe au sens de l'article 696 du code de procédure civile, devra supporter les dépens de première
instance et d'appel et sera débouté de sa demande fondée sur les dispositions de l'article 700 du code de procédure
civile.
Sur ce dernier fondement, il sera condamné à régler à la société Sogefinancement, à qui il serait inéquitable de laisser la
charge de la totalité des frais qu'elle a exposés et qui ne sont pas compris dans les dépens, une indemnité de procédure
de 1'200 euros.
Dispositif
PAR CES MOTIFS
Infirme la décision entreprise en toutes ses dispositions critiquées,
Statuant à nouveau sur les chefs infirmés et y ajoutant':
Condamne M. [F] [Y] à payer à la société Sogefinancement, pour solde du prêt de 30'000'euros souscrit le 2 février 2018,
la somme de 23'868,45'euros, majorée des intérêts au taux légal à compter du 3 septembre 2019,
Déboute M. [F] [Y] de sa demande reconventionnelle en dommages et intérêts,
Condamne M. [F] [Y] à payer à la société Sogefinancement la somme de 1'200 euros en application des dispositions de
l'article 700 du code de procédure civile,
Rejette la demande de M. [F] [Y] formée sur le même fondement,
Page 12 / 13
8 février 2024
Condamne M. [F] [Y] aux dépens de première instance et d'appel.
Arrêt signé par Madame Carole CHEGARAY, Président de la chambre commerciale à la Cour d'Appel d'ORLEANS,
présidant la collégialité et Madame Marie-Claude DONNAT , Greffier auquel la minute de la décision a été remise par le
magistrat signataire.
LE GREFFIER LE PRÉSIDENT
Page 13 / 13
|
https://www.wikidata.org/wiki/Q32295648
|
Wikidata
|
Semantic data
|
Pleias
|
CC0
| null |
Категория:Родившиеся в Приволжском районе (Астраханская область)
|
None
|
Multilingual
|
Semantic data
| 59
| 222
|
Категория:Родившиеся в Приволжском районе (Астраханская область)
категория в проекте Викимедиа
Категория:Родившиеся в Приволжском районе (Астраханская область) это частный случай понятия категория в проекте Викимедиа
Категория:Родившиеся в Приволжском районе (Астраханская область) категория объединяет темы место рождения
Категория:Родившиеся в Приволжском районе (Астраханская область) категория объединяет темы Приволжский район
Категория:Родившиеся в Приволжском районе (Астраханская область) категория содержит человек, место рождения Приволжский район
|
https://en.wikipedia.org/wiki/Ahmed%20Abdel%20Mougod%20Soliman
|
Wikipedia
|
Open Web
|
Wikimedia/Pleias
|
CC-By-SA
| 2,023
|
Ahmed Abdel Mougod Soliman
|
https://en.wikipedia.org/w/index.php?title=Ahmed Abdel Mougod Soliman&action=history
|
English
|
Written
| 56
| 97
|
Ahmed Abdel Mougod Soliman (born 19 December 1970) is an Egyptian long-distance runner. He competed in the men's marathon at the 2000 Summer Olympics.
References
1970 births
Living people
Athletes (track and field) at the 2000 Summer Olympics
Egyptian male long-distance runners
Egyptian male marathon runners
Olympic athletes for Egypt
Place of birth missing (living people)
|
uk.org.publicwhip/debate/1937-07-29a.3311.2
|
UK Hansard – House of Commons
|
Open Government
|
Pleias
|
Open Parliament Licence v3.0
| 1,937
|
Oral Answers to Questions — GOVERNMENT DEPARTMENTS. — NATIONAL DEFENCE CONTRIBU- TION (CO-OPERATIVE SOCIETIES).
|
Major Milner
|
English
|
Spoken
| 23
| 29
|
asked the Chancellor of the Exchequer whether he can indicate the estimated amount co-operative societies will pay by way of National Defence Contribution?
|
https://github.com/huntshark/validator/blob/master/test/isInteger.test.js
|
Github Open Source
|
Open Source
|
BigCode/Github/Pleias
|
MIT
| 2,018
|
validator
|
huntshark
|
JavaScript
|
Code
| 711
| 2,801
|
const isInteger = require('../src/isInteger');
const chai = require('chai');
const should = chai.should;
chai.use(require('chai-things'));
should();
describe('isInteger', function () {
// .3
it(`isInteger(.3) === false`, function () {
isInteger(.3).should.equal(false);
});
// 3
it(`isInteger(3) === true`, function () {
isInteger(3).should.equal(true);
});
// 3.
it(`isInteger(3.) === true`, function () {
isInteger(3.).should.equal(true);
});
// 3.3
it(`isInteger(3.3) === false`, function () {
isInteger(3.3).should.equal(false);
});
// '.3'
it(`isInteger('.3') === false`, function () {
isInteger('.3').should.equal(false);
});
// '3'
it(`isInteger('3') === true`, function () {
isInteger('3').should.equal(true);
});
// '3.'
it(`isInteger('3.') === false`, function () {
isInteger('3.').should.equal(false);
});
// '3.3'
it(`isInteger('3.3') === false`, function () {
isInteger('3.3 ').should.equal(false);
});
// .3
it(`isInteger(.3, {isStrict: true}) === false`, function () {
isInteger(.3, {isStrict: true}).should.equal(false);
});
// 3
it(`isInteger(3, {isStrict: true}) === true`, function () {
isInteger(3, {isStrict: true}).should.equal(true);
});
// 3.
it(`isInteger(3., {isStrict: true}) === true`, function () {
isInteger(3., {isStrict: true}).should.equal(true);
});
// 3.3
it(`isInteger(3.3, {isStrict: true}) === false`, function () {
isInteger(3.3, {isStrict: true}).should.equal(false);
});
// '.3'
it(`isInteger('.3', {isStrict: true}) === false`, function () {
isInteger('.3', {isStrict: true}).should.equal(false);
});
// '3'
it(`isInteger('3', {isStrict: true}) === false`, function () {
isInteger('3', {isStrict: true}).should.equal(false);
});
// '3.'
it(`isInteger('3.', {isStrict: true}) === false`, function () {
isInteger('3.', {isStrict: true}).should.equal(false);
});
// '3.3'
it(`isInteger('3.3', {isStrict: true}) === false`, function () {
isInteger('3.3', {isStrict: true}).should.equal(false);
});
// 0
it(`isInteger(0) === true`, function () {
isInteger(0).should.equal(true);
});
// 0.
it(`isInteger(0.) === true`, function () {
isInteger(0.).should.equal(true);
});
// .0
it(`isInteger(.0) === true`, function () {
isInteger(.0).should.equal(true);
});
// 0.0
it(`isInteger(0.0) === true`, function () {
isInteger(0.0).should.equal(true);
});
// '0'
it(`isInteger('0') === true`, function () {
isInteger('0').should.equal(true);
});
// '0.'
it(`isInteger('0.') === false`, function () {
isInteger('0.').should.equal(false);
});
// '.0'
it(`isInteger('.0') === false`, function () {
isInteger('.0').should.equal(false);
});
// '0.0'
it(`isInteger('0.0') === false`, function () {
isInteger('0.0').should.equal(false);
});
// '.'
it(`isInteger('.') === false`, function () {
isInteger('.').should.equal(false);
});
// 0
it(`isInteger(0, {isStrict: true}) === true`, function () {
isInteger(0, {isStrict: true}).should.equal(true);
});
// '0'
it(`isInteger('0', {isStrict: true}) === false`, function () {
isInteger('0', {isStrict: true}).should.equal(false);
});
// -.3
it(`isInteger(-.3) === false`, function () {
isInteger(-.3).should.equal(false);
});
// -3
it(`isInteger(-3) === true`, function () {
isInteger(-3).should.equal(true);
});
// -3.
it(`isInteger(-3.) === true`, function () {
isInteger(-3.).should.equal(true);
});
// -3.3
it(`isInteger(-3.3) === false`, function () {
isInteger(-3.3).should.equal(false);
});
// '-.3'
it(`isInteger('-.3') === false`, function () {
isInteger('-.3').should.equal(false);
});
// '-3'
it(`isInteger('-3') === true`, function () {
isInteger('-3').should.equal(true);
});
// '-3.'
it(`isInteger('-3.') === false`, function () {
isInteger('-3.').should.equal(false);
});
// '-3.3'
it(`isInteger('-3.3') === false`, function () {
isInteger('-3.3').should.equal(false);
});
// '--3'
it(`isInteger('--3') === false`, function () {
isInteger('--3').should.equal(false);
});
// '-3-3'
it(`isInteger('-3-3') === false`, function () {
isInteger('-3-3').should.equal(false);
});
// -.3
it(`isInteger(-.3, {isStrict: true}) === false`, function () {
isInteger(-.3, {isStrict: true}).should.equal(false);
});
// -3
it(`isInteger(-3, {isStrict: true}) === true`, function () {
isInteger(-3, {isStrict: true}).should.equal(true);
});
// -3.
it(`isInteger(-3., {isStrict: true}) === true`, function () {
isInteger(-3., {isStrict: true}).should.equal(true);
});
// -3.3
it(`isInteger(-3.3, {isStrict: true}) === false`, function () {
isInteger(-3.3, {isStrict: true}).should.equal(false);
});
// '-.3'
it(`isInteger('-.3', {isStrict: true}) === false`, function () {
isInteger('-.3', {isStrict: true}).should.equal(false);
});
// '-3'
it(`isInteger('-3', {isStrict: true}) === false`, function () {
isInteger('-3', {isStrict: true}).should.equal(false);
});
// '-3.'
it(`isInteger('-3.', {isStrict: true}) === false`, function () {
isInteger('-3.', {isStrict: true}).should.equal(false);
});
// '-3.3'
it(`isInteger('-3.3', {isStrict: true}) === false`, function () {
isInteger('-3.3', {isStrict: true}).should.equal(false);
});
// ''
it(`isInteger('') === false`, function () {
isInteger('').should.equal(false);
});
// ' '
it(`isInteger(' ') === false`, function () {
isInteger(' ').should.equal(false);
});
// null
it(`isInteger(null) === false`, function () {
isInteger(null).should.equal(false);
});
// undefined
it(`isInteger(undefined) === false`, function () {
isInteger(undefined).should.equal(false);
});
// NaN
it(`isInteger(NaN) === false`, function () {
isInteger(NaN).should.equal(false);
});
// Infinity
it(`isInteger(Infinity) === false`, function () {
isInteger(Number.POSITIVE_INFINITY).should.equal(false);
});
// -Infinity
it(`isInteger(-Infinity) === false`, function () {
isInteger(Number.NEGATIVE_INFINITY).should.equal(false);
});
// Object(3)
it(`isInteger(Object(3)) === true`, function () {
isInteger(Object(3)).should.equal(true);
});
// Object('3')
it(`isInteger(Object('3')) === true`, function () {
isInteger(Object('3')).should.equal(true);
});
// Object(3)
it(`isInteger(Object(3), {isStrict: true}) === true`, function () {
isInteger(Object(3), {isStrict: true}).should.equal(true);
});
// Object('3')
it(`isInteger(Object('3'), {isStrict: true}) === false`, function () {
isInteger(Object('3'), {isStrict: true}).should.equal(false);
});
// Object(-3)
it(`isInteger(Object(-3)) === true`, function () {
isInteger(Object(-3)).should.equal(true);
});
// Object('-3')
it(`isInteger(Object('-3')) === true`, function () {
isInteger(Object('-3')).should.equal(true);
});
// Object(-3)
it(`isInteger(Object(-3), {isStrict: true}) === true`, function () {
isInteger(Object(-3), {isStrict: true}).should.equal(true);
});
// Object('-3')
it(`isInteger(Object('-3'), {isStrict: true}) === false`, function () {
isInteger(Object('-3'), {isStrict: true}).should.equal(false);
});
});
|
（2016）浙0110民初字第12148号
|
Chinese-Court-Decisions
|
Open Government
|
Pleias
|
Public Domain
| 2,016
|
方清喜与浙江广诚建设有限公司建设工程施工合同纠纷一审民事裁定书
|
杭州市余杭区人民法院
|
Chinese
|
Written
| 451
| 349
|
杭州市余杭区人民法院民 事 裁 定 书（2016）浙0110民初字第12148号原告：方清喜，男，1967年2月15日出生，汉族，住浙江省淳安县。委托代理人：金星，浙江星穹律师事务所律师。委托代理人：郝碧佳，浙江星穹律师事务所律师。被告：浙江广诚建设有限公司，住所地杭州经济技术开发区杭州东部国际商务中心2幢1202室。法定代表人：杨东栋，董事长。委托代理人：练良火，浙江腾飞金鹰律师事务所律师。委托代理人：陈姣娣，浙江腾飞金鹰律师事务所律师。本院在审理原告方清喜诉被告浙江广诚建设有限公司建设工程施工合同纠纷一案，原告方清喜于2016年10月8日向本院提出撤诉申请，要求撤回对被告浙江广诚建设有限公司的起诉。本院认为，原告方清喜的撤诉申请，符合有关法律规定，据此，依照《中华人民共和国民事诉讼法》第一百四十五条第一款、第一百五十四条第一款第(五)项的规定，裁定如下：准许原告方清喜撤回起诉。本案案件受理45元（已减半），由原告方清喜负担。审判员 马 超二〇一六年十月八日书记员 张中明 来自马克数据网
|
hal-03712933-eccv2022submission.txt_1
|
French-Science-Pile
|
Open Science
|
Pleias
|
Various open science
| 2,022
|
Hierarchical Average Precision Training for Pertinent Image Retrieval. ECCV 2022, Oct 2022, Tel-Aviv, Israel. ⟨hal-03712933v2⟩
|
None
|
English
|
Written
| 7,261
| 13,171
|
HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Hierarchical Average Precision Training for Pertinent Image Retrieval
Elias Ramzi1,2, Nicolas Audebert1, Nicolas Thome1,3, Clément Rambour1 and Xavier Bitot2 1 CEDRIC, Conservatoire National des Arts et Métiers, Paris, France {elias.ramzi,nicolas.audebert,nicolas.thome,clement.rambour}@cnam.fr 2 Coexya, Paris, France xavier.bitot@coexya.eu 3 Sorbonne Université, CNRS, ISIR, F-75005 Paris, France
Abstract. Image Retrieval is commonly evaluated with Average Precision (AP) or Recall@k. Yet, those metrics, are limited to binary labels and do not take into account errors’ severity. This paper introduces a new hierarchical AP training method for pertinent image retrieval (HAPPIER). HAPPIER is based on a new H-AP metric, which leverages a concept hierarchy to refine AP by integrating errors’ importance and better evaluate rankings. To train deep models with H-AP, we carefully study the problem’s structure and design a smooth lower bound surrogate combined with a clustering loss that ensures consistent ordering. Extensive experiments on 6 datasets show that HAPPIER significantly outperforms state-of-the-art methods for hierarchical retrieval, while being on par with the latest approaches when evaluating fine-grained ranking performances. Finally, we show that HAPPIER leads to better organization of the embedding space, and prevents most severe failure cases of non-hierarchical methods. Our code is publicly available at https://github.com/elias-ramzi/HAPPIER.
Keywords: Hierarchical Image Retrieval, Hierarchical Average Precision, Ranking 1
Introduction
Image Retrieval (IR) consists in ranking images with respect to a query by decreasing order of visual similarity. IR methods are commonly evaluated using Recall@k (R@k) or Average Precision (AP). Because those metrics are non-differentiable , a rich literature exists on finding adequate surrogate loss functions to optimize them with deep learning, with tuple-wise losses [40,46,57,54,55], proxy based losses [59,53,12,49] and direct AP optimization methods [6,43,34,45,2,42]. These metrics are only defined for binary (⊕/) labels, which we denote as fine-grained labels: an image is negative as soon as it is not strictly similar to the query. Binary metrics are by design unable to take into account the severity of the mistakes in a ranking. On Fig. 1, some negative instances are “less negative” than others, e.g. given the “Brown Bear” query, “Polar bear” is more relevant than 2 E. Ramzi et al.
HAPPIER rank 1
rank 2 rank 3 rank 4 rank 5 rank
6 Query image. Baseline Fig. 1: Proposed HAPPIER framework for pertinent image retrieval. Standard ranking metrics based on binary labels, e.g. Average Precision (AP), assign the same score to the bottom and top row rankings (0.9). We introduce the H-AP metric based on non-binary labels, that takes into account mistakes’ severity. H-AP assigns a smaller score to the bottom row (0.68) than the top one (0.94). HAPPIER maximizes H-AP during training and thus
explicitly supports to learn rankings similar to the top one, in contrast to binary ranking
loss
es
.
“Butterfly”.
However
,
AP
is
0.9
for
both the top and
bottom rankings. Consequently, training on binary metrics (e.g. AP or R@k) develops no incentive to produce ranking such as the top row, and often produces rankings similar to the bottom one. To address this problem, we introduce the HAPPIER method dedicated to Hierarchical Average Precision training for Pertinent ImagE Retrieval. HAPPIER provides a smooth training objective, amenable to gradient descent, which explicitly takes into account the severity of mistakes when evaluating rankings. Our first contribution is to define a new Hierarchical AP metric (H-AP) that leverages the hierarchical tree between concepts and enables a fine weighting between errors in rankings. As shown in Fig. 1, H-AP assigns a larger score (0.94) to the top ranking than to the bottom one (0.68). We show that H-AP provides a consistent generalization of AP for the non-binary setting. We also introduce our HAPPIERF variant, giving more weights to fine-grained levels of the hierarchy. Since H-AP, like AP, is a non-differentiable metric, our second contribution is to use HAPPIER to directly optimize H-AP by gradient descent. We carefully design a smooth surrogate loss for H-AP that has strong theoretical guarantees and is an upper bound of the true loss. We then define an additional clustering loss to support having a consistency between partial and global rankings. We validate HAPPIER on six IR datasets, including three standard datasets (Stanford Online Products [33] and iNaturalist-base/full [51]), and three recent hierarchical datasets (DyML [47]). We show that, when evaluating on hierarchical metrics (e.g. H-AP), HAPPIER outperforms state-of-the-art methods for finegrained ranking [57,59,49,42], the baselines and the latest hierarchical method of [47], and only slightly under-performs vs. state-of-the-art IR methods at the fine-grained level (e.g. AP, R@1). HAPPIERF performs on par on fine-grained metrics while still outperforming fine-grained methods on hierarchical metrics. HAPPIER 2 Related work 2.1 Image Retrieval and ranking 3
The Image Retrieval community has designed several families of methods to optimize metrics such as AP and R@k. Methods that relies on tuplet-wise losses, like pair losses [17,41], triplet losses [57], or larger tuplets [46,27,54] learn comparison relations between instances. Methods using proxies have been introduced to lower the computational complexity of tuplet based training [31,59,53,12,49]: they learn jointly a deep model and weight matrix that represent proxies using a crossentropy based loss. Proxies are approximations of the original data points that should belong to their neighbourhood. Finally, there also has been large amounts of work dedicated to the direct optimization of the AP during training by introducing differentiable surrogates [6,43,34,45,2,42], so that models are optimized on the same metric they are evaluated on. However, nearly all of these methods only consider binary labels: two instances are either the same (positive) or different (negative), leading to poor performance when multiple levels of hierarchy are considered.
2.2 Hierarchical predictions and metrics
There has been a recent regain of interest in Hierarchical Classification [13,1,8] with the introductions of methods based either on a hierarchical softmax function or on multiple classifiers. It is considered that learning from hierarchical relations between labels leads to more robust models that make “better mistakes” [1]. Yet, hierarchical classification means that labels are known in advance and are identical in the train and test sets. This is called a closed set setting. However, Hierarchical Image Retrieval does not fall into this framework. Standard IR protocols consider the open set paradigm to better evaluate the generalization abilities of learned models: the retrieval task at test time pertains to labels that were not present in the train set, making classification poorly suited to IR. Meanwhile, the broader Information Retrieval community has been using datasets where documents can be more or less relevant depending on the query and the user making the request [20,24]. Instead of the mere positive/negative dichotomy, each instance has a continuous score quantifying its relevance to the query. To quantify the quality of their retrieval engine, Information Retrieval researchers have long used ranking based metrics, such as the NDCG [22,10], that penalize mistakes differently based on whether they occur at the top or the bottom of the ranking and whether wrong documents still have some marginal relevance or not. Average Precision is also used as a retrieval metric [23] and has even been given probabilistic interpretations based on how users interact with the system [14]. Several works have investigated how to optimize those metrics during the training of neural networks, e.g. using pairwise losses [4] and later using smooth surrogates of the NDCG in LambdaRank [5], SoftRank [48], ApproxNDCG [38] and LearningTo-Rank [3]. These works however focused on NDCG, the most popular metric for information retrieval, and are without any theoretical guarantees: the surrogates 4 E. Ramzi et al. are approximations of the NDCG but not lower bounds, i.e. their maximization does not imply improved performances during inference. An additional drawback of this literature is that NDCG does not relate easily to average precision [15], which is the most common metric in image retrieval. Fortunately, there have been some works done to extend AP in a graded setting where relevance between instances is not binary [44,14]. The graded Average Precision from [44] is the closest to our goal as it leverages SoftRank for direct optimization on non-binary relevance judgements, although there are significant shortcomings. There is no guarantee that the SoftRank surrogate actually minimizes the graded AP, it requires to annotate datasets with pairwise relevances which is unpractical for large scale settings and was only applied to small-scale corpora of a few thousands documents, compared to the hundred thousands of images in IR. Recently, the authors of [47] introduced three new hierarchical benchmarks datasets for image retrieval, in addition to a novel hierarchical loss CSL. CSL extends proxy-based triplet losses to the hierarchical setting and tries to structure the embedding space in a hierarchical manner. However, this method faces the same limitation as the usual triplet losses: minimizing CSL does not explicitly optimize a well-behaved hierarchical evaluation metric, e.g. H-AP. We show experimentally that our method HAPPIER significantly outperforms CSL [47] both on hierarchical metrics and AP-level evaluations.
3 HAPPIER Model
We detail HAPPIER our Hierarchical Average Precision training method for Pertinent ImagE Retrieval. We first introduce the Hierarchical Average Precision, H-AP in Sec. 3.1, that leverages a hierarchical tree (Fig. 2a) of labels. It is based on the hierarchical rank, H-rank, and evaluates rankings so that more relevant instances are ranked before less relevant ones (Fig. 2b). We then show how to directly optimize H-AP by stochastic gradient descent (SGD) using HAPPIER in Sec. 3.2. Our training objective combines a carefully designed smooth upper bound surrogate loss for LH-AP = 1 − H-AP and a clustering loss Lclust. that supports consistent rankings.
Context
Let us consider a retrieval set Ω = {xj }j∈J1;N K composed of N instances. For a query1 q ∈ Ω, we aim to order all xj ∈ Ω so that more relevant (i.e. similar) instances are ranked before relevant instances. In our hierarchical setting, the relevance of an instance xj is non-binary. We assume that we have access to a hierarchical tree defining semantic similarities between concepts as in Fig. 2a. For a query q, we leverage this knowledge to partition the set of retrieved instances into L+1 disjoint subsets Ω (l) l∈J0;LK. Ω (L) is the subset of the most similar instances to the query (i.e. fine-grained level): for L = 3 and a “Lada #2” query, Ω (3) are the images of the same “Lada #2” (green), see Fig. 2. The set Ω (l) for l < L contains instances with smaller 1 For the sake of readability, our notations are given for a single query. During training, HAPPIER optimizes our hierarchical retrieval objective by averaging several queries.
HAPPIER 5
Query Image: Lada #2 Vehicles Cars Lada Pickup Mini Bus Prius Prius #4 Lada #1 Lada #2 Lada #9 Fig. 2: HAPPIER leverages a hierarchical tree representing the semantic similarities between concepts in (a) to introduce a new hierarchical metric, H-AP in Eq. (3), see (b). H-AP exploits the hierarchy to weight rankings’ inversion: given the query image of a “Lada #2”, H-AP penalizes an inversion with a “Lada #9” less than with a “Prius #4”. To directly train models with H-AP, we carefully study the structure of the problem
and
introduce the LsH-AP loss
in
Eq. (5)
, which provides
a
smooth upper bound
of LH-AP
,
see (c).
We also
train
HAPPIER
with the
Lc
lust. loss
in Eq. (6)
to enforce the partial
ordering
in s
toch
astic
optimization to mach the global ones. relevance with respect to the query: Ω (2) in Fig. 2 is the set of “Lada” that are not “Lada #2” (blue) and Ω (1) is the set of “Cars” that are not “Lada” (orange). We also define Ω − := Ω (0) as the set of negativeSinstances, i.e. the set of vehicles L that are not “Cars” (in red) in Fig. 2 and Ω + = l=1 Ω (l). Each instance k of Ω (l) is thus associated a value through the relevance function denoted as rel(k) [20]. To rank the instances xj ∈ Ω with respect to the query q, we compute cosine similarities in an embedding space. More precisely, we extract embedding vectors using a deep neural network f parameterized by θ, vj = fθ (xj ), and compute the cosine similarity between the query and every image sj = fθ (q)T vj. Images are then ranked by decreasing cosine similarity score. We learn the parameters θ of the network with HAPPIER, our framework to directly minimize LH-AP (θ) = 1−H-AP(θ). This enforces a ranking where the instances with the highest cosine similarity scores belong to Ω (L), then Ω (L−1) etc. and the items with the lowest cosine similarity belong to Ω −.
3.1 Hierarchical Average Precision
Average Precision (AP) is the most common metric in Image Retrieval. AP evaluates a ranking in a binary setting: for a given query, each instance is either Query
Image: Lada #2
Fig. 3: Given a “Lada #2” query, the top inversion is less severe than the bottom one. Indeed on the top row instance 1 is semantically closer to the query – as it is a “Lada”– than instance 3 on the bottom row. Indeed instance 3’s closest common ancestor with the query, “Cars”, is farther in the hierarchical tree (see Fig. 2a). Because of that H-rank(2) is greater on the top row (5/3) than on the bottom row (4/3), leading to a greater H-AP in Fig. 2b for the top row. positive or negative. It is computed PN as the average of precision at each rank n over the positive set AP = |Ω1+ | n=1 Prec(n). Previous works have written the AP using the ranking operator [2] as in Eq. (1). The rank for an instance k is written as a sum of Heaviside (step) function H [38]: this counts the number of instances j ranked before k, i.e. that have a higher cosine similarity (sj > sk ). rank+ is the rank among the positive instances, i.e. restricted to Ω +. \label {eq:ap_definition} \text {AP} = \frac {1}{|\Omega ^+|} \sum _{k\in \Omega ^+} \frac {\rank ^+(k)}{\rank (k)}, \; \text {with } \begin {cases} \rank (k) = 1 + \sum _{j\in \Omega } H(s_j - s_k) \\ \rank ^+(k) = 1 + \sum _{j\in \Omega ^+} H(s_j - s_k) \end {cases} (1)
Extend
ing AP to hierarchical image retrieval
We propose an extension of AP that leverages non-binary labels. To do so, we extend the concept of rank+ to the hierarchical case with the concept of hierarchical rank, H-rank: \hrank (k) = \rel (k) + \sum _{j\in \Omega ^+} \min (\rel (k), \rel (j))\cdot H(s_j-s_k) ~. \label {eq:hierarchical_rank} (2) Intuitively, min(rel(k), rel(j)) corresponds to seeking the closest ancestor shared by instance k and j with the query in the hierarchical tree. As illustrated in Fig. 3, H-rank induces a smoother penalization for instances that do not share the same fine-grained label as the query but still share some coarser semantics, which is not the case for rank+. From H-rank in Eq. (2) we define the Hierarchical Average Precision, H-AP: \label {eq:def_hap} \hap = \frac {1}{\sum _{k\in \Omega ^+}\rel (k)} \sum _{k\in \Omega ^+} \frac {\hrank (k)}{\rank (k)} (3) Eq. (3) extends the AP to non-binary labels. We replace rank+ Pby our hierarchical rank H-rank and the normalization term |Ω + | replaced by k∈Ω + rel(k), which both represent the “sum of positives”, see more details in supplementary A.2. HAPPIER 7
H-AP extends the desirable properties of the AP. It evaluates the quality of a ranking by: i) penalizing inversions of instances that are not ranked in decreasing order of relevances with respect to the query, ii) giving stronger emphasis to inversions that occur at the top of the ranking. Finally, we can observe that, by this definition, H-AP is equal to the AP in the binary setting (L = 1). This makes H-AP a consistent generalization of AP (details in supplementary A.2).
Relevance function design
The relevance rel(k) defines how “similar” an instance k ∈ Ω (l) is to the query q. While rel(k) might be given as input in Information Retrieval datasets [37,9], we need to define it based on the hierarchical tree in our case. We want to enforce the constraint that the relevance decreases ′ when going up the tree, i.e. rel(k) > rel(k ′ ) for k ∈ Ω (l), k ′ ∈ Ω (l ) and l > l′. To do so, we assign a total weight of (l/L)α to each semantic level l, where α ∈ R+ controls the decrease rate of similarity in the tree. For example for L = 3 and α = 1, the total weights for each level are 1, 23, 13 and 0. The instance relevance rel(k) is normalized by the cardinal of Ω (l) : \label
{
eq
:hier
archy
_relevance} \rel (k) = \frac {(l/L)^\alpha }{|\Omega ^{(l)}|} \; \text {if } k \in \Omega ^{(l)} (4)
Other definitions fulfilling the decreasing similarity behaviour in the tree are possible. An interestingP option forP the relevance enables to recover a weighted L sum of AP, denoted as wAP := l=1 wl ·AP(l) ( supplementary A.2), i.e. the weighted sum of AP is a particular case of H-AP. We set α = 1 in Eq. (4) for the H-AP metric and in our main experiments. Setting α to larger values supports better performances on fine-grained levels as their relevances will relatively increase. This variant is denoted HAPPIERF and discussed in Sec. 4. 3.2
Direct optimization of H-AP H-AP
in
Eq. (3)
involv
es the computation of H-rank
and
rank
, which are nondifferentiable due to the summing of
Heaviside
step function
s
. We thus introduce a smooth approximation of H-AP to obtain a surrogate loss amenable to gradient descent, which fulfils theoretical guarantees for proper optimization
. Re-writing H-AP
In order to design our surrogate loss for LH-AP = 1−H-AP, we decompose H-rank and rank into two quantities. Denoting H-rank> (k) (resp. H-rank≤ (k)) as the restriction of H-rank to instances of strictly higher relevances (resp. lower or equal), we can see that H-rank(k) = H-rank> (k) + H-rank≤ (k). The rank can be decomposed in a similar fashion: rank(k) = rank≥ (k)+rank< (k) where < (resp. ≥) denotes the restriction to instances of strictly lower relevances (resp. higher or equal). The LH-AP can be rewritten as follow: \label {
eq:re
write
_hap} \lhap = 1 - \frac {1}{\sum _{k\in \Omega ^+}\rel (k)} \sum _{k\in \Omega ^+} \frac {\hrank ^>(k) + \hrank ^\leq (k)}{\rank ^\geq (k) + \rank ^<(k)}~. (5) 8 E. Ramzi et al. We choose to optimize over H-rank> and rank< in Eq. (5). We maximize H-rank> to enforce that the k th instance must decrease in cosine similarity score if it is ranked before another instance of higher relevance (∇H-rank> in Fig. 2 enforces the blue instance to be ranked after the green one as it is less relevant to the query). We minimize rank< to encourage the k th instance to increase in cosine similarity score if it is ranked after one or more instances of lower relevance (∇rank< in Fig. 2 enforces that the last green instance moves before less relevant instances). Optimizing both those terms leads to a decrease in LH-AP. On the other hand, we purposely do not optimize the two remaining H-rank≤ (k) and rank≥ (k) terms, since this could harm training performances as explained in supplementary A.3.
Upper bound of LH-AP
Based on the previous analysis, we now design our surrogate loss LsH-AP by introducing a smooth approximation of rank< and H-rank> (k). An important sought property of LsH-AP is that it is an upper bound of LH-AP. To this end, we approximate H-rank> (k) with a piece-wise linear function that is a lower bound of the Heaviside function. rank< is approximated with a smooth upper bound of the Heaviside that combines a piece-wise sigmoid function and an affine function, which has been shown to make the training more robust thanks to the induced implicit margins between positives and negatives [45,2,42]. More details are given in supplementary A.3 on those surrogates. Clustering constraint in HAPPIER
Positives only need to have a greater cosine similarity with the query than negatives in order to be correctly ranked. Yet, we cannot optimize the ranking on the entire datasets – and thus the true LH-AP – because of the batch-wise estimation performed in stochastic gradient descent. To mitigate this issue, we take inspiration from clustering methods [59,49] to define the following objective in order to group closely the embeddings of instances that share the same fine-grained label: \label {eq:cluster_loss} \lclust (\theta ) = - \log \left ( \frac {\exp (\frac {v_y^T p_y}{\sigma })}{\sum _{{p_z}\in \mathcal {Z}} \exp (\frac {v_y^T p_z}{\sigma })} \right ), (6) where py is the normalized proxy corresponding to the fine-grained class of the embedding vy, Z is the set of proxies, and σ is a temperature scaling parameter. In Fig. 2, ∇Lclust. further clusters “Lada #2” instances. Lclust. induces a reference shared across batches and thus enforces that the partial ordering in-between batches is consistent with the global ordering over the entire retrieval set. Our resulting final objective is a linear combination of both our losses, with a weight factor λ ∈ [0,1] that balances the two terms: \mathcal {L}_{\text {HAPPIER}}(\theta ) = (1-\lambda )\cdot \lhaps (\theta ) + \lambda \cdot \lclust (\theta ) ~. HAPPIER 4 Experiments 4.1 Experimental setup 9 Datasets
We use the standard benchmark Stanford Online Products [33] (SOP) with two levels of hierarchy (L = 2), and iNaturalist-2018 [51] with the standard splits from [2] in two settings: i) iNat-base with two levels of hierarchy (L = 2) ii) iNat-full with the full biological taxonomy composed of 7 levels (L = 7). We also evaluate on the recent dynamic metric learning (DyML) datasets (DyML-V, DyML-A, DyML-P) introduced in [47] for the task of hierarchical image retrieval, each with 3 semantic levels (L = 3).
Implementation details
Our base model is a ResNet-50 pretrained on ImageNet for SOP and iNat-base/full, and a ResNet-34 randomly initialized on DyML-V&A and pretrained on ImageNet on DyML-P, following [47]. Unless specified otherwise, all reported results are obtained with α = 1 in Eq. (4) and λ = 0.1 for LHAPPIER. We study the impact of these parameters in Sec. 4.3. Metrics For SOP and iNat, we evaluate the models based on three hierarchical metrics: H-AP – which we introduced in Eq. (3) – the Average Set Intersection (ASI) and the Normalized Discounted Cumulative Gain (NDCG), defined in supplementary B.3. We also report the AP for each semantic level. For DyML, we follow the evaluation protocols of [47] and compute AP, ASI and R@1 on each semantic scale before averaging them. We cannot compute H-AP or NDCG on those datasets as the hierarchical tree is not available on the test set. Baselines
We compare HAPPIER to several recent image retrieval methods optimized at the fine-grained level, which represent strong baselines for IR when training with binary labels: Triplet SH (TLSH ) [57], NormSoftMax (NSM) [59], ProxyNCA++ (NCA++) [49] and ROADMAP [42]. We also benchmark against hierarchical methods obtained by summing these fine-grained losses at different levels (denoted by Σ), and with respect to the recent hierarchical CSL loss [47]. Details on the experimental setup are given in supplementary B.
4.2 Main Results Hierachical results
We first evaluate HAPPIER on global hierarchical metrics. On Tab. 1, we notice that HAPPIER significantly outperforms methods trained on the fine-grained level only, with a gain on H-AP over the best performing methods of +16.1pt on SOP, +13pt on iNat-base and 12.7pt on iNat-full. HAPPIER also exhibits significant gains compared to hierarchical methods. On H-AP, HAPPIER has important gains on all datasets (e.g. +6.3pt on SOP, +4.2pt on iNat-base over the best competitor), but also on ASI and NDCG. This shows the strong generalization of the method on standard metrics. Compared to the recent CSL loss [47], we observe a consistent gain over all metrics and datasets, e.g. +6pt on H-AP, +8pt on ASI and +2.6pts on NDCG on SOP. This shows the benefits of optimizing a well-behaved hierarchical metric compared to an ad-hoc proxy method.
10
E. Ramzi et al.
Table 1: Comparison of HAPPIER on SOP and iNat-base/full when using hierarchical metrics. Best results in bold, second best underlined. SOP Hier. Fine Method iNat-base iNat-full H-AP ASI NDCG H-AP ASI NDCG H-AP ASI NDCG Triplet SH [57] NSM [59] NCA++ [49] Smooth-AP [2] ROADMAP [42] 42.2 42.8 43.0 42.9 43.3 22.4 21.1 21.5 20.6 19.1 78.8 78.3 78.4 78.2 77.9 39.5 38.0 39.5 41.3 40.3 63.7 51.6 57.0 64.2 61.0 91.5 88.9 90.1 91.9 91.2 36.1 33.3 35.3 37.2 34.7 59.2 51.7 55.7 60.1 59.6 89.8 88.2 89.0 90.1 89.5 ΣTLSH [57] ΣNSM [59] ΣNCA++ [49] CSL [47] 53.1 50.4 49.5 52.8 53.3 49.7 52.8 57.9 89.2 87.0 87.8 88.1 44.0 47.9 48.9 50.1 87.4 75.8 78.7 89.3 96.4 94.4 95.0 96.7 39.9 46.9 44.7 45.1 85.5 74.2 74.3 84.9 92.0 93.8 92.6 93.0 HAPPIER 59.4 65.9 91.5 54.3 89.3 96.9 47.9 87.2 93.8
On Tab. 2, we evaluate HAPPIER on the recent DyML benchmarks. HAPPIER again shows significant gains in mAP and ASI compared to methods only trained on fine-grained labels, e.g. +9pt in mAP and +10pt in ASI on DyML-V. HAPPIER also outperforms other hierarchical baselines: +4.8pt mAP on DyMLV, +0.9 on DyML-A and +1.8 on DyML-P. In R@1, HAPPIER performs on par with other methods on DyML-V and outperforms other hierarchical baselines by a large margin on DyML-P: 63.7 vs. 60.8 for ΣNSM. Interestingly, HAPPIER also consistently outperforms CSL [47] on its own datasets2.
Table 2: Performance comparison on Dynamic Metric Learning benchmarks [47]. Hier. Fine Method 2 DyML-Vehicle DyML-Animal DyML-Product mAP ASI R@1 mAP ASI R@1 mAP ASI R@1 TLSH [57] NSM [59] Smooth-AP [2] ROADMAP [42] 26.1 27.7 27.1 27.1 38.6 40.3 39.5 39.6 84.0 88.7 83.8 84.5 37.5 38.8 37.7 34.4 46.3 48.4 45.4 42.6 66.3 69.6 63.6 62.8 36.32 35.6 36.1 34.6 46.1 46.0 45.5 44.6 59.6 57.4 55.0 62.5 ΣTLSH [57] ΣNSM
[
59] CSL [47] 25.5 32.0 30.0 38.1 45.7 43.6 81.0 89.4 87.1 38.9 42.6 40.8 47.2 50.6 46.3 65.9 70.0 60.9 36.9 36.8 31.1 46.3 46.9 40.7 58.5 60.8 52.7 HAPPIER 37.0 49.8 89.1 43.8 50.8 68.9 38.0 47.9 63.7
CSL’s score on Tab. 2 are above those reported in [47]; personal discussions with the authors [47] validate that our results are valid for CSL, see supplementary B.5. HAPPIER 11 Detailed evaluation Tabs. 3 and 4 shows the different methods’ performances on all semantic hierarchy levels. We evaluate HAPPIER and also HAPPIERF (α > 1 for Eq. (4) in Sec. 3.1), with α = 5 on SOP and α = 3 on iNat-base/full. HAPPIER optimizes the overall hierarchical performances, while HAPPIERF is meant to be optimal at the fine-grained level while still optimizing coarser levels.
Table 3:
Comparison of HAPPIER vs. methods trained only on fine-grained labels on SOP and iNat-base. Metrics are reported for both semantic levels. SOP Hier. Fine Fine iNat-base Method R@1 AP Coarse AP R@1 Fine AP Coarse AP TLSH [57] NSM [59] NCA++ [49] Smooth-AP [2] ROADMAP [42] 79.8 81.3 81.4 81.3 82.2 59.6 61.3 61.7 61.7 62.5 14.5 13.4 13.6 13.4 12.9 66.3 70.2 67.3 67.3 69.3 33.3 37.6 37.0 35.2 35.1 51.5 38.8 44.5 53.1 50.4 CSL [47] 79.4 58.0 45.0 62.9 30.2 88.5 HAPPIER HAPPIERF 81.0 81.8 60.4 62.2 58.4 36.0 70.7 71.0 36.7 37.8 88.6 78.8
On Tab. 3, we observe that HAPPIER gives the best performances at the coarse level, with a significant boost compared to fine-grained methods, e.g. +43.9pt AP compared to the best non-hierarchical TLSH [57] on SOP. HAPPIER even outperforms the best fine-grained methods in R@1 on iNat-base, but is slightly below on SOP. HAPPIERF performs on par with the best methods at the finest level on SOP, while further improving performances on iNat-base, and still significantly outperforms fine-grained methods at the coarse level. The satisfactory behaviour and the two optimal regimes of HAPPIER and HAPPIERF are confirmed and even more pronounced on iNat-full (Tab. 4): HAPPIER gives the best results on coarser levels (from “Order”), while being very close to the best results on finer ones. HAPPIERF gives the best results at the finest levels, even outperforming very competitive fine-grained baselines. Again, note that HAPPIER outperforms CSL [47] on all semantic levels and datasets on Tabs. 3 and 4, e.g. +5pt on the fine-grained AP (“Species”) and +3pt on the coarsest AP (“Kingdom”) on Tab. 4. 4.3 HAPPIER analysis Ablation study
In Tab. 5, we study the impact of our different choices regarding the direct optimization of H-AP. The baseline method uses a sigmoid to optimize H-AP as in [38,2]. Switching to our surrogate loss LsH-AP Sec. 3.2 yields a +0.8pt increase in H-AP. Finally, the combination with Lclust. in HAPPIER results in an additional 1.3pt improvement in H-AP.
12 E. Ramzi et al
. Table 4: Comparison of HAPPIER vs. methods trained only on fine-grained labels on iNat-Full. Metrics are reported for all 7 semantic levels. Genus Family Order Class R@1 Species AP AP AP AP AP AP AP TLSH [57] NSM [59] NCA++ [49] Smooth-AP [2] ROADMAP [42] 66.3 70.2 67.3 67.3 69.3 33.3 37.6 37.0 35.2 35.1 34.2 38.0 37.9 36.3 35.4 32.3 31.4 33.0 33.5 29.3 35.4 28.6 32.3 35.0 29.6 48.5 36.6 41.9 49.3 46.4 54.6 43.9 48.4 55.8 54.7 68.4 63.0 66.1 69.9 69.5 CSL [47] 59.9 30.4 32.4 36.2 50.7 81.0 87.4 91.3 HAPPIER HAPPIERF 70.2 70.8 36.0 37.6 37.0 38.2 38.0 38.8 51.9 50.9 81.3 76.1 89.1 82.2 94.4 83.1 Hier. Fine Method Phylum Kingdom Table 5: Impact of optimization choices Table 6: Comparison of H-AP (Eq. (4)) for H-AP (cf. Sec. 3.2) on iNat-base. and ΣwAP from supplementary A.2. LsH-AP Lclust. H-AP ✗ ✓ ✓ ✗ ✗ ✓ 52.3 53.1 54.3 test→ train↓ H-AP P wAP NDCG H-AP P wAP 53.1 52.0 39.8 40.5 97.0 96.4 Impact of
the relevan
ce function
Tab. 6 compares P models that are trained with the relevance function of Eq. (4), i.e. H-AP, andP wAP (relevance given in supplementary A.2). We report results for H-AP, wAP and NDCG. Both P H-AP, wAP perform better when trained withP their own metric: +1.1pt H-AP for the model trained to optimize it and +0.7pt wAP for the model trained to optimize it. Both models show similar performances in NDCG (96.4 vs. 97.0). 54 37.6 52
H-AP AP
fine 53 37.4 37.2 51 50 37.0 49
APfine
36.8
1.0 1.5 2.0 3.0 4.0 5.0 H-AP 48 0.0 0.1 0.2 0.3 0.4 0.5 0.7 α λ (a) APfine vs α in Eq. (4). (b) H-AP vs. λ for LHAPPIER. Fig. 4: Impact on Inat-base of α in Eq. (4) for setting the relevance of H-AP (a) and of the λ hyper-parameter on HAPPIER results (b).
HAPPIER 13 Hyper-parameters
Fig. 4a studies the impact of α for setting the relevance in Eq. (4): increasing α improves the performances of the AP at the fine-grained level on iNat-base, as expected. We also show in Fig. 4b the impact of λ weighting LsH-AP and Lclust. in HAPPIER performances: we observe a stable increase in H-AP within 0 < λ < 0.5 compared to optimizing only LsH-AP, while a drop in performance is observed for λ > 0.5. This shows the complementarity of LsH-AP and Lclust., and how, when combined, HAPPIER reaches its best performance.
4.4 Qualitative study
We provide here qualitative assessments of HAPPIER, including embedding space analysis and visualiz ation of HAPPIER’s retrievals. t-SNE: organization of the embedding space
In Fig. 5, we plot using tSNE [50,7] how HAPPIER learns an embedding space on SOP (L = 2) that is well-organized. We plot the mean vector of each fine-grained class and we assign the color based on the coarse level. We show on Fig. 5a the t-SNE visualisation obtained using a baseline method trained on the fine-grained labels, and in Fig. 5b we plot the t-SNE of the embedding space of a model trained with HAPPIER. We cannot observe any clear clusters for the coarse level on Fig. 5a, whereas we can appreciate the the quality of the hierarchical clusters formed on Fig. 5b. Controlled errors Finally, we showcase in Fig. 6 errors of HAPPIER vs. a fine-grained baseline. On Fig. 6a, we illustrate how a model trained with HAPPIER makes mistakes that are less severe than a baseline model trained only on the fine-grained level. On Fig. 6b, we show an example where both models fail to retrieve the correct fine-grained instances, however the model trained with HAPPIER retrieves images of bikes that are visually more similar to the query. (a) t-SNE visualization of a model (b) t-SNE visualization of a model trained only on the fine-grained labels. trained with HAPPIER.
Fig. 5: t-SNE visualisation of the embedding space of two models trained on SOP. Each point is the average embedding of each fine-grained label (object instance) and the colors represent coarse labels (object category, e.g. bike, coffee maker). rank 2 rank 3 rank 4 rank 5 rank 6 HAPPIER
rank 1
Baseline Query image
(a) HAPPIER can help make less severe mistakes. The inversion on the bottom row are with negative instances (in red), where as with HAPPIER (top row) inversions are with instances sharing the same coarse label “bike” (in orange).
rank 2 rank 3 rank 4 rank 5 rank 6 HAPPIER rank 1 Baseline Query image
(b) In this example, the models fail to retrieve the correct fine grained images. However HAPPIER still retrieves images of very similar bikes (in orange) whereas the baseline retrieves images that are dissimilar semantically to the query (in red).
Fig. 6: Qualitative examples of failure cases from a standard fine-grained model corrected by training with HAPPIER.
5 Conclusion
In this work, we introduce HAPPIER, a new training method that leverages hierarchical relations between concepts to learn robust rankings. HAPPIER is based on a new metric H-AP that evaluates hierarchical rankings and uses a combination of a smooth upper bound surrogate with theoretical guarantees and a clustering loss to directly optimize it. Extensive experiments show that HAPPIER performs on par to state-of-the-art image retrieval methods on fine-grained metrics and exhibits large improvements vs. recent hierarchical methods on hierarchical metrics. Learning more robust rankings reduces the severity of ranking errors, and is qualitatively related to a better organization of the embedding space with HAPPIER. Future works include the adaptation of HAPPIER to the unsupervised setting, e.g. for providing a relevant self-training criterion.
Acknowledgement HAPPIER 15
References 1. Bertinetto, L., Mueller, R., Tertikas, K., Samangooei, S., Lord, N.A.: Making better mistakes: Leveraging class hierarchies with deep networks. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 12506–12515 (2020) 3 2. Brown, A., Xie, W., Kalogeiton, V., Zisserman, A.: Smooth-ap: Smoothing the path towards large-scale image retrieval. In: European Conference on Computer Vision. pp. 677–694. Springer (2020) 1, 3, 6, 8, 9, 10, 11, 12, 28, 29 3. Bruch, S., Zoghi, M., Bendersky, M., Najork, M.: Revisiting approximate metric optimization in the age of deep neural networks. In: Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. pp. 1241–1244 (2019) 3 4. Burges, C., Shaked, T., Renshaw, E., Lazier, A., Deeds, M., Hamilton, N., Hullender, G.: Learning to rank using gradient descent. In: Proceedings of the 22nd International Conference on Machine Learning. p. 89–96. ICML ’05, Association for Computing Machinery, New York, NY, USA (2005). https://doi.org/10.1145/1102351.1102363, https://doi.org/10.1145/1102351.1102363 3 5. Burges, C., Ragno, R., Le, Q.: Learning to rank with nonsmooth cost functions. In: Schölkopf, B., Platt, J., Hoffman, T. (eds.) Advances in Neural Information Processing Systems. vol. 19. MIT Press (2006), https://proceedings.neurips. cc/paper/2006/file/af44c4c56f385c43f2529f9b1b018f6a-Paper.pdf 3 6. Cakir, F., He, K., Xia, X., Kulis, B., Sclaroff, S.: Deep
learning to rank. 7. Chan, D.M., Rao, R., Huang, F., Canny, J.F.: Gpu accelerated t-distributed stochastic neighbor embedding. Journal of Parallel and Distributed Computing 131, 1–13 (2019) 13, 31 8. Chang, D., Pang, K., Zheng, Y., Ma, Z., Song, Y.Z., Guo, J.: Your” flamingo” is my” bird”: Fine-grained, or not. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 11476–11485 (2021) 3, 28 9. Chapelle, O., Chang, Y.: Yahoo! learning to rank challenge overview. In: Proceedings of the learning to rank challenge. pp. 1–24. PMLR (2011) 7 10. Croft, W.B., Metzler, D., Strohman, T.: Search engines: Information retrieval in practice, vol. 520. Addison-Wesley Reading (2010) 3, 31 11. Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: Imagenet: A large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition. pp. 248–255 (2009). https://doi.org/10.1109/CVPR.2009.5206848 28 12. Deng, J., Guo, J., Xue, N., Zafeiriou, S.: Arcface: Additive angular margin loss for deep face recognition. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 4690–4699 (2019) 1, 3 13. Dhall, A., Makarova, A., Ganea, O., Pavllo, D., Greeff, M., Krause, A.: Hierarchical image classification using entailment cone embeddings. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. pp. –837 (2020) 3 14. Dupret, G., Piwowarski, B.: A user behavior model for average precision and its generalization to graded judgments. In: Proceedings of the 33rd International ACM SIGIR Conference on Research and Development in Information Retrieval. p. 531–538. SIGIR ’10, Association for Computing Machinery, New York, NY, USA (2010). https://doi.org/10.1145/1835449.1835538, https://doi.org/10.1145/1835449.1835538 3, 4 16 E. Ramzi et al. 15. Dupret, G., Piwowarski, B.: Model based comparison of discounted cumulative gain and average precision. Journal of Discrete Algorithms 18, 49–62 (2013). https://doi.org/https://doi.org/10.1016/j.jda.2012.10.002, https://www.sciencedirect.com/science/article/pii/S1570866712001372, selected papers from the 18th International Symposium on String Processing and Information Retrieval (SPIRE 2011) 4 16. Fagin, R., Kumar, R., Sivakumar, D.: Comparing top k lists. SIAM Journal on discrete mathematics 17(1), 134–160 (2003) 30 17. Hadsell, R., Chopra, S., LeCun, Y.: Dimensionality reduction by learning an invariant mapping. In: 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’06). vol. 2, pp. 1735–1742. IEEE (2006) 3 18. Harris, C.R., Millman, K.J., van der Walt, S.J., Gommers, R., Virtanen, P., Cournapeau, D., Wieser, E., Taylor, J., Berg, S., Smith, N.J., Kern, R., Picus, M., Hoyer, S., van Kerkwijk, M.H., Brett, M., Haldane, A., del R’ıo, J.F., Wiebe, M., Peterson, P., G’erard-Marchant, P., Sheppard, K., Reddy, T., Weckesser, W., Abbasi, H., Gohlke, C., Oliphant, T.E.: Array programming with NumPy. Nature 585(7825), 357–362 (Sep 2020). https://doi.org/10.1038/s41586-020-2649-2, https://doi.org/10.1038/s41586-020-2649-2 31 19. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. 20. Hjørland, B.: The foundation of the concept of relevance. Journal of the American Society for Information Science and Technology 61(2), 217–237 (2010) 3, 5 21. Hunter, J.D.: Matplotlib: A 2d graphics environment. Computing in Science & Engineering 9(3), 90–95 (2007). https://doi.org/10.1109/MCSE.2007.55 31 22. Järvelin, K., Kekäläinen, J.: Cumulated gain-based evaluation of ir techniques. ACM Transactions on Information Systems (TOIS) 20(4), 422–446 (2002) 3 23. Järvelin, K., Kekäläinen, J.: Ir evaluation methods for retrieving highly relevant documents. In: ACM SIGIR Forum. vol. 51, pp. 243–250. ACM New York, NY, USA (2017) 3 24. Kekäläinen, J., Järvelin, K.: Using graded relevance assessments in ir evaluation. Journal of the American Society for Information Science and Technology 53(13), 1120–1129 (2002). https://doi.org/https://doi.org/10.1002/asi.10137, https://onlinelibrary.wiley.com/doi/abs/10.1002/asi.10137 3 25. Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014) 30 26. Krause, J., Stark, M., Deng, J., Fei-Fei, L.: 3d object representations for fine-grained categorization. In: 4th International IEEE Workshop on 3D Representation and Recognition (3dRR-13). Sydney, Australia (2013) 27 27. Law, M.T., Thome, N., Cord, M.: Learning a distance metric from relative comparisons between quadruplets of images. International Journal of Computer Vision 121(1), 65–94 (2017) 3 28. Liu, Z., Luo, P., Qiu, S., Wang, X., Tang, X.: Deepfashion: Powering robust clothes recognition and retrieval with rich annotations. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (June
) 27
29. Micikevicius, P., Narang, S., Alben, J., Diamos, G., Elsen, E., Garcia, D., Ginsburg, B., Houston, M., Kuchaiev, O., Venkatesh, G., et al.: Mixed precision training. arXiv preprint arXiv:1710.03740 (2017) 31 30. Miller, G.A.: Wordnet: A lexical database for english. Commun. ACM 38(11), 39–41 (nov 1995). https://doi.org/10.1145/219717.219748, https: //doi.org/10.1145/219717.219748 28 HAPPIER 17 31. Movshovitz-Attias, Y., Toshev, A., Leung, T.K., Ioffe, S., Singh, S.: No fuss distance metric learning using proxies. In: Proceedings of the IEEE International Conference on Computer Vision. pp. 360–368 (2017) 3 32. Musgrave, K., Belongie, S., Lim, S.N.: Pytorch metric learning (2020) 31 33. Oh Song, H., Xiang, Y., Jegelka, S., Savarese, S.: Deep metric learning via lifted structured feature embedding. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 4004–4012 (2016) 2, 9, 27, 28 34. P., M.V., Paulus, A., Musil, V., Martius, G., Rolı́nek, M.: Differentiation of blackbox combinatorial solvers. In: ICLR (2020) 1, 3 35. Parikh, D., Grauman, K.: Relative attributes. 41. Radenović, F., Tolias, G., Chum, O.: Cnn image retrieval learns from bow: Unsupervised fine-tuning with hard examples. In: Leibe, B., Matas, J., Sebe, N., Welling, M. (eds.) Computer Vision – ECCV 2016. pp. 3–20. Springer International Publishing, Cham (2016) 3 42. Ramzi, E., Thome, N., Rambour, C., Audebert, N., Bitot, X.: Robust and decomposable average precision for image retrieval. Advances in Neural Information Processing Systems 34 (2021) 1, 2, 3, 8, 9, 10, 11, 12, 27, 28, 31 43. Revaud, J., Almazán, J., Rezende, R.S., Souza, C.R.d.: Learning with average precision: Training image retrieval with a listwise loss. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 5107–5116 (2019) 1, 3 44. Robertson, S.E., Kanoulas, E., Yilmaz, E.: Extending average precision to graded relevance judgments. In: Proceedings of the 33rd international ACM SIGIR conference on Research and development in information retrieval. pp. 603–610 (2010) 4 45. Rolı́nek, M., Musil, V., Paulus, A., Vlastelica, M., Michaelis, C., Martius, G.: Optimizing rank-based metrics with blackbox differentiation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 7620–7630 (2020) 1, 3, 8 46. Sohn, K.: Improved deep metric learning with multi-class n- loss objective. In: Lee, D., Sugiyama, M., Luxburg, U., Guyon, I., Garnett, R. (eds.) 18 E. Ramzi et al. Advances in Neural Information Processing Systems. vol. 29. Curran Associates, Inc. (2016), https://proceedings.neurips.cc/paper/2016/file/ 6b180037abbebea991d8b1232f8a8ca9-Paper.pdf 1, 3 47. Sun, Y., Zhu, Y., Zhang, Y., Zheng, P., Qiu, X., Zhang, C., Wei, Y.: Dynamic metric learning: Towards a scalable metric space to accommodate multiple semantic scales. 48. Taylor, M., Guiver, J., Robertson, S., Minka, T.: Softrank: Optimizing non-smooth rank metrics. In: Proceedings of the 2008 International Conference on Web Search and Data Mining. p. 77–86. WSDM ’08, Association for Computing Machinery, New York, NY, USA (2008). https://doi.org/10.1145/1341531.1341544, https://doi.org/10.1145/1341531.1341544 3 49. Teh, E.W., DeVries, T., Taylor, G.W.: Proxynca++: Revisiting and revitalizing proxy neighborhood component analysis. In: European Conference on Computer Vision. pp. 448–464. Springer (2020) 1, 2, 3, 8, 9, 10, 11, 12, 31 50. van der Maaten, L., Hinton, G.: Visualizing high-dimensional data using t-sne. Journal of Machine Learning Research 9, 2579–2605 (2008) 13 51. Van Horn, G., Mac Aodha, O., Song, Y., Cui, Y., Sun, C., Shepard, A., Adam, H., Perona, P., Belongie, S.: The inaturalist species classification and detection dataset. In: Proceedings of the IEEE conference on computer vision and pattern recognition.
|
US-1905283258-A_1
|
USPTO
|
Open Government
|
Baber
|
Public Domain
| 1,905
|
None
|
None
|
English
|
Written
| 2,099
| 2,948
|
Vapor electric apparatus.
PATENTED MAR. 24, 1908.
0. 0. KRUH.
VAPOR ELECTRIC APPARATUS.
APPLICATION FILED OGT.1B 1905.
2 SHBETSr-SHEET 1.
Inventor Osias O. Kruh,
y fitty.
Witnesses:
Fig.
Fig. 3.
Witnesses PATENTBD MAR. 24, 1908- 0. 0. KRUH.
VAPOR ELECTRIC APPARATUS.
APPLIOATION nun 0011a. 1006.
2 sums-sum z.
Osias O. Kruh,
UNITED sTArns r ENT OFFICE.
OSIAS O. KRUH, OI SCHENEC'IADY, NEW YORK, ASSIGNOR TO GENERAL ELECTRICCOMPANY, A CORPORATION OF NEW YORK.
' VAPOR ELECTRIC APPARATUS.
Specification of Letters Patent.
Patented March 24, 1908.
Application filed October 18, 1905. Serial No. 283,258.
To all whom it may concern:
Be it known that I, OSIAS O. linen, a subject of the Emperor ofAustria-Hungary, residing at Schenectady, county of Schenectady, andState of New York, have invented certain new and useful Improvements inVapor Electric Apparatus, of which the following is a specification.
My resent invention relates to apparatus in which electric discharges orarcs take place in evacuated inclosures, and more particularly toapparatus using alternating current as a source of current supply.
The invention may be embodied in devices of this character used asrectifiers, lamps or the like, and embodies various novel features ofconstruction as pointed out more particularly inthe appended claims.
The invention itsell, however, will be better understood by reference tothe following description taken in connection with the accompanyingdrawings in which Figure 1 represents one embodiment of my invention;Fig. 2 a detail, and Figs. 3 and 4 modified forms of my invention.
In vapor electric devices, so called because the arcs therein aremaintained through the medium of a vapor, a plurality of anodes areusually employed where the apparatus is to be used in connection. with asource of alternating current sup ly. These anodes constitute thetermina s of electric arcs which play between the res ective anodes anda common ne ativc e ectrode or cathode.
Under certam conditions of operation it.
sometimes happens that an arc takes place from anode to anode, therebyforming a short circuit, which though momentary may be sullicientlyviolent, or if often repeated, to injure or destroy the a paratus. lovercome this trouble by provning the anodes with shields which directthe arc stream emanating from the cathode in a manner to protect theanodes from any s )attering or other emission of nercury from thecathode, of which material the cathode is usually composed, and whichserve also to protect the anodes from any other radiation or emanationfrom the cz-rthodc. The shields are also arranged to protect the anodesfrom. any dropping of globules oi condensed mercury-taking place in theevacuated inclosure in which the anodes are located.
In Fig. 1, I have represented a mercury vapor rectifier for changingalternating current into direct current. The rectifiertube consists ofan exhausted receptacle 1, generally of glass, which "tube is providedwith suitable electrodes. The positive electrodes or anodes areindicated at 2 and 3 and may be formed of cylinders or plates, or othersuitable shapes, of artificial graphite, titanium carbid, iron, or thelike. Current is conveyed to each electrode by means of leadingin wireswhich pass through the wall ofthe receptacle 1. hi the case of the anode2, the leading-in wire, usually of platinum, is shown at 4 and is sealedthrough the inner end of a reentrant portion 5 of the tube 1. Theportion of the wire 4 extending between the anode and the reentrantmember 5 is 'surrounded by a coating of vitreous material 5, such aslava, hard glass, or the like. The shieldstor the anode 2 may, ofcourse, vary widely in form and still retain their essential functions.The present arrangement, however, consists of an inner shield 6 and anouter shield 7. This shield 6 is a tube of glass or the like, the lowerend of which may be sealed about the part 5, while the upper end extendsover and beyond the anode. 2 and is closed. An opening 8, however, isformed in the side of the tubular shield 6, which serves as a passagefor the are stream going to the anode 2. The outer shield '7 surroundsthe shield 6 and is conveniently arranged concentric therewith. Thisshield is closed at the top as indicated, and may be.
sealed at its bottom portion about the part 5 by which it, as well asthe inner shield 6 'is supported. An o'ening 9 into the outer shleld 7is provider near the lower end'there of and preferably at the side 0posite, the anode 3 which cooperates with t e anode 2. The anode3 isprovided with shields 10 and 11 substantially the same in constructionand arrangement as the shields 6 and 7. Thus no special description ofthese shields is necessary.
The negative electrode or cathode with which the anodes 2 and 3cooperates, con sists of a body of mercury 12 located in a pocket formedin the lower end of the tube 1. A leading-in wire 13 serves as usual toconvey current flowing between this electrode and an outside circuit. Astarting anode, which like the electrode 12 is of mercury, is providedfor starting the apparatus into operation in the usual manner. ing intothe small pocket or chamber con- The opentaining the mercuryconstituting the starting electrode, is shown in Fig. 1 'at 14, whilethe de ending portion of the pocket is indicate ingdottedlines. Thisstarting electrode and the pocket containing it is, however, shownbetter by'the detail view in Fig.
rectly to alternating current supply mains such as 16 and 17. Inductancecoils 18 and 19 are joined in series across these mains. The junctionpoint 20 between them we tends to a consumption circuit 21, the otherterminal of which is connected to the oathode 12. For the purpose ofstarting the rectifier the starting anode 13 is connected through aresistance or other current limiting device 22 to one of the supplyleads.
By shaking or tiltingthe apparatus so as to bring the mercury of theelectrodes 12 and 15. into momentary contact, a starting arc ensueswhich when the current is in the proper direction, causes the main arcsto start between the cathode and the main anodes 2 and 3. A number, oftrials may perhaps benecessary in some cases before the apparatusstarts, as the direction of current may not at the'first trial be suchas required. When the apparatus is in operation the arc stream from thecathode 12goes to each of the anodes 2 and 3 and before reaching thecorresponding anode it is constrained to follow a path which bends onitself. Thus the arc stream from the cathode 12 passesinto the opening 9and then up through the tubular assage afforded between the concentricwal s of the shields 6 and 7 and then into the opening 8 in the innershield 6 through which it passes to the anode 2. The arc stream as itpasses to the opening 9 is of the nature of a high velocity blast whichnecessarily carries along with it more or less imperfectly vaporizedmercury. The greater portion of this superfluous matter is thrown off bythe arc stream as it turns into the opening 9. As the arc streamcontinues farther and then again changes its direction to pass into theopening 8 any remaining superfluous mercury is likewise projected beyondthe opening by its momentum. The anode is thereby effectually protectedfrom any spattering or other emanation from the cathode. Moreover it isrotected also from any falling drops or globu es of mercury which maydescend from the walls of the tube 1, as the mercury vapor generated inthe apparatus is condensed. Also the anode is protected from the effectof any radiation, perhaps in the nature of ultra violet rays, emanatingfrom the cathode surface. This protection is afforded by the interposedwalls of the two glass shields 6 and 7. The whole arrangement thusconstitutes a most effective means for preventing arcing between anodes.
I have indicated in passing, some of the causes which I believe areactive in producing arcing between anodes. I do not wish to beconsidered as guaranteeing the accuracy of what from my observations Inow consider to be the most likely causes of the arcing, since thebeneficial effects of my in vention may be obtained, regardless of thereasons therefor, by following the modes of construction which I havepointed out.
Instead of using the apparatus shown in Fig. 1 I may use an arrangementsomewhat as shown in Fig. 3. In this figure I have shown merely therectifier tube without indicating the connections of such tube to itselectric circuits, as such connections are well understood. In Fig. 3,the evacuated receptacle is indicated at 23. The graphite anodes areshown at 24 and 25 and the cooperating cathode at 26. The anodes are asbefore surrounded by cylinders 26 and 27 closed at both ends andsupported in any suitable manner as from the leading-in conductors ofthe respective anodes. Communication with the anode is afforded in thecase of the cylinder 26 by an opening 27 formed in the upper portion ofthe cylindrical wall and on that side of the cylinder away from theanode 25. A cylindrical shield 28, closed at the top and open at thebottom, is mounted concentrically over the inner shield 26 and may beconveniently supported from a small standard 29 extending from theclosed top of the inner shield 26 as indicated. The other anode 25 islikewise provided with an outer shield 30 mounted and arranged in thesame manner as the shield for the anode 24, so that no furtherdescription thereof is necessary. In either case the outer shields28-and 30, in addition to serving as an additional safe guard to protectthe anodes against contact with falling or otherwise projected particlesof mercury, serves also to direct the arcstream passing from the cathodeto the anode. The walls of the shield 28 for example constrain the arcstream to pass in a direction tangential tothe inner shield 26 andthereby prevents. superfluous matter, and particularly condensedmercury, from entering the opening 27 into the inner shield 26surrounding the anode 24. The momentum acquired by the particles as theyare carried along in the arc stream causes them to be projected past theopening 27 so that only that portion of the arc stream enters which isnecessary as a vehicle for conveying current.
Still another arrangement of the protect-- ing shields may be employedas indicated in Fig. 4. In this figure the evacuated receptacle of therectifier is indicated at 31.
I by a number of openings such as at 36, 37
and the like, located about the sides of the shield as indicated.
The inner shield 35 is closed at its top and open at the bottom andsurrounds the anode 32. It may conveniently be supported by a small rod38 of glass or the like depending from the inner side of the top of theshield 34. The cooperating anode 33 is provided with similarlyconstructed and arranged shields 39 and which re uire no specialdescription. The are from t e cathode 34 enters the openings 36, 37 andthe like, then passes down between the walls of the shields 34 and 35and then up through the open bottom of the shield 35 to the anode 32.The are stream is thereby effectually freed from any superfluous matterand the danger from arcing, due to contact'of such matter with theanode, is thus obviated. The. anode moreover is also well protectedagainst contact with the mercury condensing and falling Within therectifier tube.
The devices shown in this application are covered broadly by the claimsin my application Serial No. 194,520 filed February 20,
\ 1904, which is now involved in interferences. This present caseislimited to'the improvements on the invention of said formerapplication which are herein described and.
claimed, and the claims herein are therefore not to be construed ascovering anything disclosed in my said rior ap lication.
It is evident t at various modifications may be made in .the embodimentof my invention without departing from the spirit thereof, for whichreason I do not wish to be limited to the details shown and. described.
What I claim as new and desire to secure by Letters Patent'of the UnitedStates, is
1. The combination of a receptacle or container, electrodes therein, andshields therefor arranged one inside of the other.
2. The combination of a receptacle or container, electrodes therein, andshields for said electrodes having non-registering openings therein.
3. The combination of a rece tacle or container, a vaporizable electrodet erein, a plu rality of cooperating electrodes, and shields for each ofsaid cooperating electrodes, the shields for each electrode beinglocated one inside the other.
4. The combinationof a rece tacle or container, electrodes therein, ands ields for one at least of said electrodes, said shields being arrangedone about the other and rovided with openings located to produce a cange in direction of the arc stream going to the electrode.
In witnesswhereof, I have hereunto set my hand this'17'th day ofOctober, 1905.
' OSIAS O. K'RUH. Witnesses:
BENJAMIN B. HULL, HELEN ORFORD.
|
https://github.com/voteflux/members.flux.party/blob/master/src/components/UserRevocation.vue
|
Github Open Source
|
Open Source
|
BigCode/Github/Pleias
|
MIT
| null |
members.flux.party
|
voteflux
|
Vue
|
Code
| 336
| 1,211
|
<template>
<UiSection title="Revoke your Membership" :dangerZone="true">
<Warning>
This will remove all your data from our database.<br>
To undo this you'll need to sign up again.<br>
There is no going back.
</Warning>
<error v-if="this.err.revoke.msg">
{{ this.err.revoke.msg }}
</error>
{{ this.err.revoke.msg }}
<transition name="fade" mode="out-in" class="mv3">
<router-link :to="R.Dashboard" v-show="state != DONE">Take me back to safety.</router-link>
</transition>
<div class="mt4">
<h3>Revocation</h3>
<transition name="fade" mode="out-in">
<div v-if="state == AT_START" :key="AT_START">
<button class="danger-btn db" v-on:click="startRevocation()">Revoke my Membership</button>
</div>
<div v-else-if="state == CONFIMATION" :key="CONFIMATION">
<h4>Please confirm by filling out:</h4>
<label class="db">Your Email ({{ user.email }})</label><br>
<input class="db" type="email" v-model="formEmail" placeholder="confirm your email"/>
<button class="mt2 danger-btn db" :disabled="formEmail !== user.email" v-on:click="confirmRevocation()">Revoke my Membership</button>
</div>
<div v-else-if="state == CONFIRMATION_2" :key="CONFIRMATION_2">
<h4>Last Step:</h4>
<button class="mt3 danger-btn db" v-on:click="doRevocationFinally()">Revoke my Membership</button>
<button class="mt4 db" v-on:click="cancelRevocation()">Cancel - Take me to safety</button>
</div>
<div v-else-if="state == SAVING" :key="SAVING">
<h4>Revoking membership...</h4>
</div>
<div v-else-if="state == DONE" :key="DONE">
<h4>Your membership has been revoked. You should recieve a confirmation email.</h4>
<button class="db mt2" v-on:click="checkMembership()">Okay</button>
</div>
</transition>
</div>
</UiSection>
</template>
<script lang="ts">
import Vue from "vue";
import { UiSection } from "./common";
import { mkErrContainer } from "../lib/errors";
import { Error, Warning } from "./common";
import { M, MsgBus } from "../messages";
import R from "../routes";
enum Cs {
AT_START,
CONFIMATION,
CONFIRMATION_2,
SAVING,
DONE,
}
export default Vue.extend({
components: { UiSection, Error, Warning },
props: ["auth", "user"],
data: () => ({
err: mkErrContainer(),
state: Cs.AT_START,
formEmail: "",
R,
...Cs
}),
methods: {
init() {
this.formEmail = ""
this.state = Cs.AT_START
},
startRevocation() {
this.formEmail = ""
this.state = Cs.CONFIMATION
},
confirmRevocation() {
this.state = Cs.CONFIRMATION_2
},
doRevocationFinally() {
this.state = Cs.SAVING
this.$flux.v1.revokeMembership(this.$props.user)
.then((r) => r.do({
failed: (e, errObj) => {
if (errObj.status == 403) {
this.err.revoke = this.$err("Unauthorized... Have you already revoked your membership?", e)
} else {
this.err.revoke = this.$unknownErr(e)
}
},
success: () => this.state = Cs.DONE
}))
},
cancelRevocation() {
this.init()
},
checkMembership() {
MsgBus.$emit(M.REFRESH_USER)
}
},
created() {
}
});
</script>
<style lang="scss" scoped>
@import "tachyons";
.danger-btn {
@extend .bg-dark-red;
@extend .white;
}
.danger-btn:disabled {
background-color: #6300007c;
}
</style>
|
https://arz.wikipedia.org/wiki/%D8%A8%D9%8A%D8%AA%D8%B1%20%D9%85%D8%A7%D8%AA%D9%8A%D8%A7%D8%B3
|
Wikipedia
|
Open Web
|
Wikimedia/Pleias
|
CC-By-SA
| 2,023
|
بيتر ماتياس
|
https://arz.wikipedia.org/w/index.php?title=بيتر ماتياس&action=history
|
Egyptian Arabic
|
Written
| 58
| 188
|
بيتر ماتياس كان مؤرخ من المملكه المتحده.
حياته
بيتر ماتياس من مواليد يوم 10 يناير سنة 1928.
الدراسه
درس فى Colston's School و مدرسه بريستول للقواعد.
العضويه
كان عضو فى:
الاكاديميه الاوروبيه
جوايز
زميل الاكاديميه البريطانيه
نيشان الامبراطوريه البريطانيه من رتبه قائد
وفاته
بيتر ماتياس مات يوم 1 مارس سنة 2016.
لينكات
مصادر
مؤرخين
مؤرخين من المملكه المتحده
|
https://dadosabertos.web.stj.jus.br/stj/267371978
|
Superior Tribunal de Justiça
|
Open Government
|
Pleias
|
Public Domain
| null | null |
Portuguese
|
Written
| 553
| 1,179
|
DECISÃO
Cuida-se de agravo interposto por WALESKA GONCALVES DOS SANTOS CINTRA, contra decisão que inadmitiu recurso especial.
É, no essencial, o relatório. Decido.
Mediante análise do recurso de WALESKA GONCALVES DOS SANTOS CINTRA, verifica-se que incide o óbice da Súmula n. 284/STF, uma vez que não houve a indicação do permissivo constitucional autorizador do recurso especial, aplicando-se, por conseguinte, a referida súmula: "É inadmissível o recurso extraordinário, quando a deficiência na sua fundamentação não permitir a exata compreensão da controvérsia".
Isso porque, conforme disposto no art. 1.029, II, do CPC/2015, a petição do recurso especial deve conter a "demonstração do cabimento do recurso interposto".
Sendo assim, a parte recorrente deve evidenciar de forma explícita e específica que seu recurso está fundamentado no art. 105, inciso III, da Constituição Federal, e quais são as alíneas desse permissivo constitucional que servem de base para a sua interposição.
Esse entendimento possui respaldo em recente julgado desta Corte:
PROCESSO CIVIL. ADMINISTRATIVO. IMPROBIDADE ADMINISTRATIVA. INCIDÊNCIA POR ANALOGIA DO ENUNCIADO N. 284 DA SÚMULA DO STF. DEFICIÊNCIA RECURSAL. ART. 1.029 DO CPC/2015. AUSÊNCIA DE INDICAÇÃO DE DISPOSITIVO VIOLADO. PRETENSÃO DE REEXAME FÁTICO-PROBATÓRIO. INCIDÊNCIA DO ENUNCIADO N. 7 DA SÚMULA DO STJ.
..
II - Na espécie, incide o óbice da Súmula n. 284/STF, uma vez que não houve a correta indicação do permissivo constitucional autorizador do recurso especial, aplicando-se, por conseguinte, a referida Súmula: "É inadmissível o recurso extraordinário, quando a deficiência na sua fundamentação não permitir a exata compreensão da controvérsia".
III - Conforme disposto no art. 1.029, II, do CPC/2015, a petição do recurso especial deve conter a "demonstração do cabimento do recurso interposto". Sendo assim, o recorrente, na petição de interposição, deve evidenciar de forma explícita e específica em qual ou quais dos permissivos constitucionais está fundado o seu recurso especial, com a expressa indicação da alínea do dispositivo autorizador. Este entendimento possui respaldo em antiga jurisprudência desta Corte Superior de Justiça, que assim definiu: "O recurso, para ter acesso à sua apreciação neste Tribunal, deve indicar, quando da sua interposição, expressamente, o dispositivo e alínea que autoriza sua admissão.
.. .
(AgInt no AREsp n. 1.479.509/SP, relator Ministro Francisco Falcão, Segunda Turma, DJe de 22/11/2019.)
Confiram-se ainda os seguintes julgados: AgInt no AgInt no AREsp n. 1.015.487/RJ, relator Ministro Marco Aurélio Bellizze, Terceira Turma, DJe de 2/8/2017; AgRg nos EDcl no AREsp n. 604.337/RJ, relator Ministro Ericson Maranho (desembargador convocado do TJ/SP), Sexta Turma, DJe de 11/5/2015; e AgRg no AREsp n. 165.022/SP, relator Ministro Marco Aurélio Bellizze, Quinta Turma, DJe de 3/9/2013; AgRg no Ag 205.379/SP, relator Ministro José Delgado, Primeira Turma, DJ de 29/3/1999; AgInt no AREsp n. 1.824.850/MG, relator Ministro Marco Aurélio Bellizze, Terceira turma, DJe de 21/06/2021; AgInt no AREsp n. 1.776.348/SP, relator Ministro Moura Ribeiro, Terceira Turma, DJe de 11/06/2021.
Caso exista nos autos prévia fixação de honorários advocatícios pelas instâncias de origem, determino sua majoração em desfavor da parte recorrente, no importe de 15% sobre o valor já arbitrado, nos termos do art. 85, § 11, do Código de Processo Civil, observados, se aplicáveis, os limites percentuais previstos nos §§ 2º e 3º do referido dispositivo legal, bem como eventual concessão da gratuidade da justiça.
Ante o exposto, com base no art. 21-E, V, do Regimento Interno do Superior Tribunal de Justiça, não conheço do recurso.
Publique-se. Intimem-se.
EMENTA
|
|
greekenglishlexi0000henr_w1t1_75
|
English-PD
|
Open Culture
|
Pleias
|
Public Domain
| 1,846
|
Greek-english lexicon based on the german work of francis passow
|
Henry George Liddell, M.A., and Robert Scott, M.A
|
English
|
Written
| 7,677
| 16,728
|
10, 222; also with past, Att., Matth., Gr. § 524, 3.—3. with optat., followed by subj., with ay, Il. 11, 386; in Att. this use is dub.—4. the first clause with dy is left out, when it can be easily supplied from the context. 3, 52:9, 245, etc.; or its place is supplied by a part., Ul. 10, 246 —é. 350; 15, 213. EI press a wish, If only... O that... would that... Il. 24, 74, Herm. Vig. n. 190; but ei@e, ef yép and ai ydép are more free. v. ef ydp.—6. with Optat., as a sort of particle of time, of repeated actions, as often as, whenever, Thuc. 7, 79, usu. with impf. or plqpf., sometimes with aor.—II. wirH INDIC., where possibility is asserted, without expressing any uncertainty or question; if, since.—1. with indic. pres., ef pw’ Eoger ToAEuilew, GAAovE wev ners or, Il. 3,67, where no doubt is thrown on the supposition.—2. with ‘indic. past, esp. in oaths and prayers, ef NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39, etc., v. el- toTe.—3. with indic. fut NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39, etc., v. el- toTe.—3. with indic. fut NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39, etc., v. el- toTe.—3. with indic. fut NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39, etc., v. el- toTe.—3. with indic. fut NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39, etc., v. el- toTe.—3. with indic. fut NOTE ToL éxi vndv Epewa, THdE [0b Kpynvov &éAdwp Il. 1, 39 , yydceat, el Kai Beorecin TOALY ob GAaTaésetc, Il. 2, 367,379, where the fut. is looked on ascertain: Att. the optat: with av freq. follows, to soften the positiveness of the phrase, Soph. El. 244. So the indic. often follows, even after the opt. expressing a simple supposed case, e. g. weot pevoiveor, el TeAéovorr, Il. 12,59, they tried whether they could; where they are represented as it were saying, We will try whether we can, O as to add vivacity to the sentence: esp. oft. in Att. Prose. The indic. pres. or fut. is also put after ef in protasis, when not a mere probability, but a necessary result on a condition is intended, Il. 5, In Att., e¢ with indic. is used not only of probable, but of actual events, to qualify the positive assertion, and so much like 671: most freq. after Gavudlo, also after other verbs, esp. expressing strong feeling, e. g. dyavaxtéw, detvov ToLoduat, dndoi, ete., Hdt. 1, 155, Thuc. 6, 60, Plat. Lach. 194 A.—4. In Att. where ei with impf. is followed by dy with impf., the first implies that a condition has not been fulfilled, the second that a result has therefore not taken place; e. g. ef Te elyev, édidov dy, if he had it, he would give it... (but he has it not.)—5. with indic. aor., followed by indic. aor. with dv, it expresses the same thing in reference to a past time, for which in Lat. both verbs would have been in subj. plapf., ef rz dy, Edwxev dv had he had it, he would have given it, cf. Il. 21, 211, 544. In this case the impf. with dv may follow, ei érreicOnv, ok dv Hppdcrovr, had 1 obeyed, I should not have been ill, Buttm. Gramm. § 139, 9, 4. and 10: sometimes, but not often, this day is left out with the impf. 7v, Thuc. 1, 37. More rarely the opt. with day follows ei wf and the indic. aor., Il. 5, 388 ;17, 70.—I. wirH suBsUNCT., ei is scarcely to be distinguished from édv, though an attempt has been made to explain ef as expressing greater probability in the condition, suppose that, Kihner Ausf. Gr. § 818, Anm. 1, Herm. Soph. Ant. 706 ; much more rare than the former, but most freq. in Hom., Il. 1, 340, Od. 5, 221, etc.: ef kev with subjunct. being the more freq. For the Att. it was for- merly laid down that only éév or 77v, never ef was used with subjunct. : but many exceptions are found in Trag., as Soph. O. T. 198, 874, O. C. 1443, Ant. 710, 1032, cf. Herm. Aj. 491: also in comic wr., as Ar. Eq. 698, 700, Pac. 450.: nay it has been admitted even in prose, as Thuc. 6, 21, Xen. Mem. 2,.1, 12, Plat. Phaedr. 234, Rep. 579 E: in later‘authors «/ veth subjunct. is very common, with cptat., without apodosis, # «- | BH rm. Vig. n. 304: cf. also ef Ke— EIAP IV. with ParTicip. instead of in fie where oti is usu. supplied, but rave Soph. Aj. 686, and Herm. ib. 179, Bornem. Xen. Mem. 2, 6, 25.—V. WITH INFIN., sometimes in Hat. e g. 3, 105, 108, in orat. obliqua.— Ey from the first clause must sometimes be supplied with each of several fol- lowing clauses, even when these are in different moods, Schaf. Mel. p. 111. Whether, in indirect questions and after verbs containing a question, doubt, uncertainty, etc., or Bede, etc., we do not know whether to be a god, Il. 5, 183; in Hom. also free in ellipt. clauses, where wepoevoc, okorov, etc., must be supplied, etc. KnpdKecot KéAEvoaY, Gui TPL OTi out Tpinoda péyay, (TELPNOGLEVvOL) & memiOotey Ilndeidnv, trying whether they could move Achilles, Il. 23, 40; where the optat. without dv is used, because the action is past; ef. fl. 10, 206; 20, 464; if present or future, it would require e¢ xe or ééy with subj., Il. 5, 279, though Att. ef with subj. is used even in this sign. — C. Regularly ei begins the sentence, and so is followed by the articles: hence all compds., as & Keé, el mp, el uh, et Kal, etc., may, be best referred to their own specific heads. It is preceded by one or two conjunctions: — I. «ai el and if, even though, implying that the case is not so, Il. 20, 371; Kai et mov, Od. 7, 320, also kai ef ke, which follows the same rules as ef ce, Att. cel, xaév, Kavei: in Att. also 6uwe is oft. added in ap odosis (even though, yet still), though this word is sometimes attached to the end of the conditional clause, to which it adds force, Aesch. Pers. 295, Cho. 115: care must be taken not to confuse the end with ed, Herm. Vig. n. 307.—II. ob? ei, nay not if, not even if, Il. 5, 645; 20, 102, Od. 4, 293. —III. ed. and ei Te or (as Wolf writes it) cel, ooel Te, as if, as though, in comparisons, Od. 7, 36, I, 13, 492, 19, 366, Od. 19, 39: the Att. also inserts dy or wep, Gorep él, O¢ av el, Gorep dv ei or Worepavel, Heind. Plat. Gorg. 479 A. Ei, Dor. for 7 and of, cf. mez. Eid, also properisp. ela, and poet. trisyll. éia, Lat. eia, a cheering or stimulating exclamation, on! up! away! Trag., etc.: also come on then! Aesch. Ag. 1650, and Plat. ; ela vuv, well now! Ar. Pac. 459, stronger than dye vuy: also ela Oy: éa and etaare akin to it. [@ al- ways, whence Gramm. wrote éia, Vv. Reisig de Constr. Antistr. p. 19.] Eig, 3 sing. imperf. act. from édo, Hom. ; Eidew, f. -dow, to cry sia, like aidGw from ai, and ebdfw from eda, v, Valck. Diatr. p. 20. Eiduevy, ie, 7. & low, moist pasture, water-meadow, év eiaweva EAeoc, Il. 4. 483, in Ap. Rh. a flooded meadow. (Usu. deriv. from efatat, jvTat, neat, juevoc, Whence some Gramm. wrote ciapevy, cf. xéOnuwac: Buttm, however, V. #ldecc, connects it with Hiov.) ; Ei dv, Ep. and Ion. e ke Contr. into ééy and jy. But et...dv seems permissible both in Hom., and Att., where some words come between, Il. 2, 597, cf Herm. Vig. n. 303, Elavoc, 7, 6v, Ep. for éaéc, Il. 16,9. Elapéuacboc, ov, (slap, pacbdc\ with youthful, sweetly breasts, Ant. 397 Eldpo- span, &c, (etap, TépToual) joying in spring, Orph. Eldpo- span, &c, (etap, TépToual) joying in spring, Hom., Elatat, eiaro, 3 pl. pres. and impf., poet. for Ion. Earaz, gato, and this for qvraz, WTO, from 7juat, Hom. filato, 3 plur. imperf. mid. from et‘, for #vTo, i.e. #oav, occurs only Ou. 20, 106, where Buttm. Ausf. Gr. Anm. 14, n., would read ei-aro. Efaro, 3 sing. plqpf. mid. from év- pout for elvto, they had on. E/Bipoc, ov, trickling: from EI’BQ, Ep. form of AeiBw, to drop, let fall in drops, Hom., who regul. uses it in phrase daxpvoy eiBecy and Kata Odxpvov elBewv, to shed tears. Mid. to trickle or run down, drip, Hes. ‘Th. 210: but also as in act., ddxpva elBouévn, Soph. Ant. 527. Ei ydp, for if..., Il. 20, 26: but usu. —II. expressing a wish, O 7f..., O that..., would that... Lat. utimam! c. optat., ed yap ’A@nvn doin Kapto¢ éuoé, Il. 17, 561, so ef yap tot, Od. 17, 513, and ef yap mwc, Od. 16, 148. But Hom. ig tage has ai ydp, ai vip On, at yap 6n ToTE, at yap Two. Tike (Moraine: use c. inf. is rare, at yap, Too &WV;...240¢ yauBpo¢g Ka- A€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€€ wish without alluding to its result, Nitzsch Od. 1, 265. Elye, if however, adding a condition which makes the thing dependent upon it unlikely or impossible, e.g., oikévee e0éAetc (évat etye wév el- detc, 6ooa Tol aica Kyde’ dvaTAn- out, évOade k’ ade wév tode OGua ov 4ococc, thou wishest to go home: yet of thou didst know..., ete., Od. 5, 206,—II.. if then, since, Lat. siquidem, of things which are taken for granted, Il. 1, 393, Od. 16, 300, Herm. Vig. n. 310. Cf. etzep. Ei yodv, even if, implying that the thing will not be so, only once Hom., viz. Il. 5; 258, ubi al. ef y ody. Ei 0’, dye, used in cheering, etc., on then, come on! oft. in Hom., who also has plur., e/ 0’, dyere, Il. 22, 381. He usu. joins ei. 0’, dye viv, el 0’; aye On, et. 0’, dye unv, or el 0, dye tot, followed by imperat., also ef 0’, tye Tol, Od. 9, 37. For the imperat. dedpo is, found, Il. 17, 685, and in speaking to one’s self the subj. aor., Od. 9, 37, or indic. fut., Il. 1, 524; 9, 167. The phrase is elliptic, and would be in full ef &’ é0¢Aece or ef dé BobAet, kye, but if thou wishest, come, and so serves to qualify the imperat., like Lat. sis vide, fac sis, agite sultis, Nitzsch Od. 1, 270. Eidaévouat, poet. lengthd. form of eldomat, to be like, rwvi, Nic. ElddAmoc, n, ov, (eidog) formed; here.ce shapely, comely, Od. 24, 279.— [L. like, looking like, Anth. ElddAAouat, = eldaivouat, ivdda- AQUAL. ElddAAouat, = eldaivouat, ivdda- AQUAL. ElddAAouat, arog, 76, (&du, as if lengthd. poet. from &dap) food, meat, victuals, Hom.—2 Of cattle, fodder, forage, etc. 5, 369.—3. also a bait for fish, Od. 12, 252.—4. weAicone dvOiuov eidap, of honey-cakes, Theocr. 15, 115. Ep. word, Ei 0é, with no apodosis, is elliptic, as Il. 9, 46, ef dé kal adrol, devydv-ran, but a eer (will), let them flee, where é0éAovcr is to be supplied, as in ef 0’, dye: so too 9, 262, e dé, (0€Aetc), a ee pev Gxovoov. In Il. 21, 487, and Od. 2, 115, the apodosis is implied in the protasis.—II. in complete sentences, but if, even if, oft. in Hom. It may be followed by any particle which follows e/, v. esp. El wéy: on ei 0’ od and ef 0’ ovv V. ec un. We have the notion of et dé strengthd. in ed 0’ av, if on the other hand, Od. 16, 105. Eidéa, ac, 7, for idéa, dub. in Ar. Thesm. 438. Eidéyera, opt., and eidévaz, inf. of olda, q. V. Ei 02 pH, v. sub et 7}. Eidéyera, ac, #, an odious, ugly look, LX X.: from EldéyOae, é¢, (eldoc, &yBoc) of hateful look, in genl. ugly, Polyb.: putrid, fetid, Hipp. Eidéw, for eid, subj. from oida. Ki 67, expressing a supposition which cannot be contradicted, 7f now, seeing that, ll. 1, 61, esp. after 7, Il. 1, 294, 574: also in Indirect questions, whether now, Od. 1, 207: always c. indicat. Eidjua, aroc, 76, (eidévar) knowledge. Eidnovixdc, ady., with knowledge, skilfully. Eidnovix, ov, gen. ovoc, (eldévat) knowing, experienced, skilled, espé, tivéc, Clem, Ad. Adv. -uévaec. Eidnoév, Ep. inf. fut. for eidévat, of eidévat, of eidévat. Eidotc, ewe. 7, a thing, Strab.: from Eidorolée, 6v, (eldoc, rotéw) specific, characteristic of a species. Eldoé, coc, 76, (*eldw) that which is seen, the form, shape, figure, Lat. species; freq. of human form in Hom., who usu. has the acc. eldoé dptoroc, ayntoc, Kakdc, GAéyKtog, buoLoe, ete.; sometimes opp. to the understanding, sometimes to bodily strength, v. Od. 17, 454, Il. 21, 316: also of the appearance, look, as of a dog, Od. 17, 308, cf. déuac. Esp. beautiful form, like Lat. forma, Hdt. 1, 199; 8, 105, ete. In Trag. periphr. for the person, Soph. El. 1177.—Il. in genl., a form, figure, fashion, sort, particular kind, cidea, Tov KvBwv, Hdt.'1, 94, eidoc vdcov, Thue. 2, 50, etce.: esp. species, opp. to yévoc, genus, hence also = idéa, Plat., and Arist., cf. Ritter Hist. of Philos. 2, 265, sqq.—III. in later authors ra edn are spices, fine and costly wares. Eldodopéw, O, (eldoc, gépw) to represent, express, Dion. H. EidvAatov, ac, 7, Eidyia, wife of Aescus, Lyc. 1024. EidvAatov, ov, 76, dim. from eldoc strictly a@ little form or image: usu. @ short, highly wrought, descriptive poem, mostly, but by no means only, on pastoral subjects, an idyll, cf. Plin. Ep. 4, 14. EidvAouat,= eiddA2ouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat,= eiddA2ouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. EidvAouat, = eiddAouat, eidaivo pat, Pemp. ap. Stob. p. 46, 9. E A. to see, behold, look at, mostly in aor. eidov, in Hom. and Ep. oft. without augm. Zdov, inf. idezy, in I. and Ep. also idwui, part. idov in Hom. freq. with an adv.,{+~. a,dvra, dypetov lidar, eyeing with astern glance, etc.: he also freq. has more fully 6¢0arpoiorv id. The same act. signf. belongs to the aor. mid. eidéunv, in Hom. more freq. Ep. idwut, inf. idécOaz, subj. idwuat, imperat. idod: with which Hom. has also 6¢6a2,uoi-ovv, or more freq. év 666.,-to see before the eyes: this tense alone is joined with recpdouat, in phrase aye, Te phoouat 768 idwuat, well, I will make trial and see, Od. 6, 126, cf. 21, 159: also without zecpdouar; Just our te look and see, Od. 4, 22; 10, 44. But Hom. also uses both aorists of mental sight, to see, perceive, as must be the case in I. 21,61, dd9a idwyar evi dpe-civ, 708 daeta, ef. ll. 4, 249, Od. 21; 112. This definiteness belongs only to the oldest Greek: in later poets to perceive by any of the senses, Jac. A. P. p. 189. In construction, idea and idéfae are either absol., or used c. acc. followed by a relative clause, where the relative is to be resolved by 67z, so that the “acc. is not strictly the object, but belongs to the verb in the relat. clause, e. g. Od. 10, 195, eldov..vicov, tHv mépt movToc éoteddvura:, i. e. eidov éTt TEept vyoov ovToc éoreduivwtat; ‘though in the remarkable passage, GAbyxov...odTt yap Ide, he saw, i. e. enjoyed not the favor of his spouse, Il. 11, 243, ydéprv is the object; (this phrase must not be confounded with xapv eidéva, v. infr.): freq. also ideiv & Tt, more rare éxé Te, Il. 23, 143, and mpéc Tu, Od. 12, 244, to look at or towards a thing. The imperat. mid. (dod, see, occurring first in Att., is mostly used as an exclamation, lo! behold! Lat. ecce: but it is then written (dod, or sometimes idoy. where it is a true imperat. it remains idod, e.g. idod we, Eur. Hec. 808. Opdw is used as pres., épaxa as perf., dwouat as fut. (for eldfow belongs to signf. B, to know.) But to the signf. to see, belong—II. the Ep. and Ton. pass. and mid. eldouac: aor elodunv, in Hom. also éevodunv, ao, arc, In pass. signf., to be seen, appear, seem, eideTat Hap, doTpa, the day, the stars are visible, appear, Il. 8, 555, cf. 24, 319, Od. 5, 283: metaph., 76 Oé tor Kp elderar elvar, for that seems unto thee to be very death, that is very death in thine eyes, Il.-1, 228, cf. Od..9, 11, etc.; and feq aira "Oyce Képdlov eicato Oud: hence— Z. to have the appearance or look of a thing, take the appearance, make a show of a thing, eica7’ iwev é¢ Ajuvor, he made a show of going to Lemnos, Od. 8, 283; eloato, Oc te furor, it had the look as of a shield, Od. 5, 281; and c. dat., to make one’s self like, be like, eicato foyynv TloAirn, she made herself like Polites in voice, Il. 2, 791, cf 20, 81. Most usu. in part. pres. and aor., eidéuevoc, eioduevoc, fetodjtevoc, besides which Hom. uses only 3 sing. pres. and aor.; and once 2 sing. and 3 plur. aor. An impf. eels he was seen, occurs first in Ap. B. to know: which sign. comes from the perf., for what one has seen or observed, that one knows: hence the word is mostly used of mediate knowledge, whilst for such as is immediate, cavalleria is most usu., Wolf Dem. 461, 2. The tenses which belong to this sign are these: perf. used as pres., oida (in Alcae. 94 e, p. 72, dida) I know, c. part. eidde, inf. eidévae, Ep. tduevac and iduev, imperat. icf. sabj. eidd, Ep. also idéw, opt. eideiny: plapf. as imperf. noe and jdea, Att.-ndy, I knew: fut. eicouat, more rarely and mostly Ep. eidjow (also in Hat. 7, 234): aor. and perf..are supplied from yeyvd- oxw: though in later Greek we have an aor. eiéjaat, Arist. Magn. Mor. 1, 1,3, tc. The forms are so irreg. in pres. ard impf, that they can only be fullyicated of in grammars. In Hom., ica, and Dor., oida¢ is 2 sing. perf. for vic@a, e G. Od. 1, 337, (in Att. also sometimes oticac, Cratin. Malth. 10, cf. Meineke Menand. p. 122: iduev 1 pl. for icwev: besides iSuevar and tduev, inf. for eidévar: idéa subj. for eld, Il. 14, 235. eidere 2 pl. subj. for eld, Od. 9, 17, eidere uev for eldduer’, Il. 1, 363, and idvia fem. part. for eldvia, but only in phrase idviqat zparidecat: plapf. 2 and 3 sing. 7eidyc, jeidn for ydne, Hon. Il. 22, 280, Od. 9, 206, 3 pl. tcav for yaar, IL 18, 405, Od. 4, 772; yaar, Eur. Cyel. 231. n both futures, yet e/djow only in Il. 1, 546, Ep. int. eligoeuev, Od. 6, 237, where it almost passes Into signf. A, to see, and so in the hymns. For the rest v. Buttm. Ausf. Gr. § 199, 111, and Catal. in voc. [icacx has usu. f, as Od. 2, 211, but sometimes also j, in arsis, as Od. 2, 283.] -In Hom. it must be rendered sometimes by to know, understand, have knowledge of, sometimes by to know, discern, perceive; later to come to know, learn; though it may be so taken however in Od. 2, 16: very freq. strenethd. by ed or odga, esp. ev _ olda, I know well, and Part. et eiddc, also eb icOc, know well, be assured. It is often followed by a clause with cic, éxwe or 6rt, and, in case of doubt, with eZ, whether, rarely with the relat. Also followed by acc., or in-fin. Hom. has the peculiar usage, vuata, pndea olde, he is knowing, skilled in wise counsels; and so still more free with ads., memvuuéva, Keyaplouéva, gira, UpTla, iTLa, Ked-ya, GOeuiorra eidévat, but usu. in nart. e{ddac. In this sign, to be skilled im, the word also takes a gen. in Hom., mostly indeed c. part., e. g. rozuv ev eiddc, cunning with the pets ol TEKTOOUVaWY, wayNC, ELC.; but #: in pres. indic., it 15, 412. The inoperat. is freq. in protestations. Lastly Hom. uses, EIEN like totw Zetc, icrw vir Zeve, let Jove know it, be witness, Hom.; Dor. ittw Z.: ydpiv eidévat tevi, to acknowledge a debt to another, thank him, first in Il. 14, 235, Hat. 3, 21, but most freq. in Att., and prose. Post-Hom., usages:—1. to be in a condition, be able, have the power, c. inf., Jac. Anth. 2, 1, p. 308—2. oid’ 671, oic@ dre, used absol. parenthetically as a particle of affirmation, J know, ou know it well, Wolf Dem, 508, 17, eind. Plat. Gorg. 486 B.—3. olc@’ ody; freq. interrog. form, usu. answered by oo’ olda, Valck. Hipp. 598. —4. olc#’ 671, also oicf’ 6 and oic’ @¢, followed by imperat., gives a command without specifying what, as if this was known before, esp. oia@’ 6 dpadcov, for dpdcov, oic@’ 6, v. sub dpdo. (The word always has the digamma in Hom., Fidov, Fede Ete., which remains in Lat. videre, Sanskrit. vid scire, Germ. wissen, our to wit or wot. On the difference of eidwaver from ywyvékerr, V. ywyvéoka, fin. EidwAeiov, ov, 76, (eidwAdrov, bbw) sacrificed to idocs; as subst. 70 eid., Nas EidwAoAarpeia, ac, 7, worship of idols, idolatry, N.T.: and EidwrodarTpéw, G, to worship idols, Eccl.: from EidwAoAdrpnc, ov, 6, 7, (eidwAoAov, | Aarpec) an idol-worshipper, idolater. Eido26u0pdoc, ov, (eidwAov, wopdy) formed after a likeness, like an image, Geop. e EidwAov,ov, 76, (eidoc) ashape, figure, image: in Hom. of disembodied spirits, esp. GpoTdv eidwha KkauovTwy : any unsubstantial form, esp. a vision, phantom, Hom., etc.: hence a phantom of the mind. a fancy, Plat. Phaed. 66 C. —Il. an image in the mind, idea, Xen. Symp. 4, 21: esp. with the Stoics, Cic. Fam. 15, 16.—Il. an image, statue, yuva.xoc, Hat. 1, 51, 6, 58—2. esp. of a god; hence an idol, false god, LXX.—IV. eidwaa odpavia, the constellations, Lat. signa, Ap. Rh. Eidwiorhacréw, 0, to form, model, Heracl. + from EidwiérAactoc, ov, (eidwAoréc) to make an image, eldwAov eid., Plat. Rep. 605 C: to represent by an image or figure, tid, Diod.—2. to body, image forth, depict by words, Longin. Hence y Eidwhoroinate, ewc, 7, « figuring : representation, Sext. imp EidwAoroiéc, ac, 7,=foreg., Plat. Tim. 46 A. EidwAoroiéc, 4, dv, (eidwdo- totoc) of, belonging to figuring or re- presenting, Tévn, Plat.-Soph. 235 A. j sb mad OV; oe ToLéw) figuring, forming, making figures or weer: as subst. 6 e/d., Plat. Soph. 239 D. Eidwrovpyicdc, 9, 6v, (eidwrov, *%nyw)=cldwAoroikoc, Plat. Soph. 266 D. Eldwrovpyicdc, é, (etdwhov, puivo- feat) like an image, Plut.? Eldwrovpyc, é, (eldwrov, yaipw) delighting in idols, Synes. led Elev, Att. 3 plur. opt. from efué, for eijcay, be it so, well, good, proceed, or to proceed, Lat. esto: a very common particle, esp. in Att. dialogue, in passing to the next point, Herm. Eur. Supp. 795: the phrases GAA elev, elév ye, clev Of are more rare. also to express impatience, Ar, Nub. 176. [elev in Att. poets is sometimes used as a spondee, Aesch. Cho. 657, Ar. Pac. 663.] Elevator, opt. aor. 2 act. from type: but eyv, opt. pres. from sip.4 Elevator, adv., (elevate) at once “orth- with, instantly, Al, and Ion. Elevator, inter. J wish! O that! would that! Lat. wtinam! Od. 2, 33: the Dor. ai#e is more free. in Homer: on all OgeAAov and ddeAov, ec, e, ¥. dgeiAw: Cc. opt., of things possible, but not likely; with the past tenses of indic. of things impossible: later also the inf. follows e/@e, Herm. Vig. n. 190, a, cf. sub ed yap. Elevator, E. -iow, poet. for 26iw Elevator, perf. pass. from é6iGw, in the accustomed manner Diog. L. Elevator, Att. for govxa, q. v. Elevator, perf. from the. “ Elevator, ov, 6, (eikac) epith of the Epicureans, because they commemorated their founder’s death. On the twentieth of Gamelion, Ath. 298 D. Eixalo, f.-dow, Att. pass. 7xa-cua, Dind. Ar. Eq. 230, Piers. Moer. p. 182, and on. the eugm. in genl. v. Buttm. Ausf. Gr. § 84, Anm. 3, (e/xéc). To make like to, represent by an image or likeness, portray, Xen. Oec. 10, 1. hence in pass., eleov ypady eixacpe-vy, a figure coloured to the life, Hat. 2, 182; aleroc eixaou., a figure like an eagle, Id. 3, 28; hence—IlI. to liken, compare, Ti Ttvt, Aesch. Cho. 633; ek. TL kai Tl, Hdt. 9, 34, etc.: hence to compare and infer something, to conjecture, guess, Lat. conjicere, esp. im phrase @¢ elxaoat, Hat. 2, 104, et3.: and c. dupl. acc. to guess to be, [Lut. 4, 31, Aesch. Supp. 288, Soph. Ant. Pass. to be like, resemble, revi Eur. Bacch. 942, 1253; also apug tu va, Ar. Ach. 783. Elevacy, inf. of a lengthd. aor. ei Kabov, from etka, to yield, Soph. ete, ; for there is no such pres. es eikdbw, Elmsl. Med. 186, Ellendt Lex. Soph in v. Ei Kai, even though, although, c. indic., Hom. ; c. op é., Il.: distinguished from «ai el by expressing that the thing is really so, Herm, Vig. n. 307: ef. ei C. ElkatoBovalu, ac,7,rashness, Ecel.: from EixaPcvAoe, ov, (elkatoc, Bovan- rash, ill-edwised, Eccl. Eixa:cA5yoc, ov, (eikatoc, Aéyo, talking ut random, Philodem. ap. Vol Herevl. 2, 10. Firacouv0éw, 6, to speak inconsiderately; and Kixaouidia, ac, h, thoughtless talking, ususeless babble; from Elkar6uddoc, ov, (sixaioc, wiboc, talking at random or to no purpose, Eccl. 5 Hikacoppnuoovvn, G,=elkavopvbew, Eixaroppnuoovvn, n¢ = elkacouv- Gia: from é Elxacoppuorv, ov, gen. ov, (elkat oc, pyua)=elixacouvboc. Bixaioc, ata, aiov, without plan purpose: random, rash, hasty, nearly = Lat. temerarius, Soph. Fr. 288.— IL—rvyeév, casual, hence common, worthless, Luc. Adv. -wc, Joseph Hence. Eixavoctvn ne, 7, thoughtlessness Timon ap. Diog. L, 5 11. 399 EIKO Bixarétys, ntoc, 7, =fcreg., Philo- | warring against images, assaulting im- dcm. ap. Vol. Hercul. 2. 9. Eikdc, Gdoc, 7, (elkoor) the number twenty, for elxotde.—Il. the twentieth day of the month, sub. 7uépa, Hes. Op. 790, 818: also pl. eladdec, Ar. Nub. 17. One of the days of the Eleusinian mysteries was also so called, Eur. Ion 1076. Eixdoa 1 aor. inf. act. from ela- Q. “Hixdodw, Aeol. and Dor. for elxa- $a, Sapph. 34. Wixdoxia, ac, 7, (elkafw) a likeness, image, representation, Xen.—II. a comparison, Plut.: a conjecture, a guess-ing, Plat. Rep. 534 A.: Rixacua, atoc, 76, (elxdlw) a likeness, image, Aesch. Theb. 523. Eixaopoc, od, 6, (elkalw) one who conjectures, a guesser, diviner, TOV weA-A6vrwy, Thue. 1, 138. Hixaorexde, 4, 6v, (eixdla) of, belonging, suited to representing, guessing, OF interpreting: 7 eik., Sub. Téyvn, the art of copying or portraying, Plat. Soph. 235 D, ete.: rd eix., sub. ér(bpyuura, aduerbs of doubting. Adv. ‘KOC, by conjecture, by guessing, Elkaarée, %, bv, (e1xagw) to be compared, like, Soph. T. 699: copied, represented, i; Eixdre, Dor. for elxoae. Elke, <i sev, and ¢t...dv, if, very freq. in Hom., and Ep., the same as say, q. V., usu. c. subj, but. c. opt. Od. 7, 315: Att. c. Opt., never.c. subj., Plat. Legg. 807 C, Xen. Ages. 1, 1, cf. Matth. Gr. Gr. § 525, 7, a. Onits difference from ai «xe v. Thiersch Gr. Gr. § 327, cf. § 329, 330. Elkscy plqpf. act. of tue. Like Averpoc, ov, (etkedoc, dverpoc) dream-like, Ar, Av. 687. Elxedoc, 7, ov, (elxdc) like, after the form or fashion of, tv Hom.: also see aoe Eixg, Adv. of eixatoc, without plan or purpose, heedlessly, rashly, at random, at venture, Lat. temere, Hipp., Tragg., Plat., etc.—2. in vain, to no purpose, N. 'T. Rom. 13, 4. HixoBoréa, G(eikh, BoAj, BdAAw) to aim or act at random, at a venture, Ar. Pr. 549, Hixovica, f. -low, (eixdv) to mould, fashion, Plut.:. to copy. Elkovirde, 7, Ov, (elxov) representing a figure, copied from it, dyahud Tvo, a portrait statue, Callix. ap. Ath, 205 F.—II. counterfeited, forged, pretended, Anth. Adv. -Kdc. Eikéviov, ov, 76, dim. from elxdy, a little image or figure, Polem. ap. Ath. 574 C. Elxévicue, atog, 76, (eixovitw) a topy, image, Anth. Hixovicwog, od, 0, (elkovilw) a delineation, esp. by words, Lat. effictio, Plut. = Eixovoypavéw, @, to delineate, describe, Philo; and Elxovoypidia, ac, 9, @ sketch, description, Strab.: from Elkovoypadog, ov, (elkav ypddw) painting figures; as subst., 4 painter, Arist. Poet. [a] Hixovoacyia, ae, 7, (etxov, Adyoc) aapeatins speaking, Plat. Phaedr. 267 Eixovoudyia, ac, 4, a war against dols or images, Eccl. : from ? oS) ¢ BMeOMOUCY 9G, ov, (elkor, udyouat) ages, Eccl. : from ? oS) ¢ Elxovorolé6c, 6v, (éixév, Totéw) making figures or images As subst., Arist. Poet. Elxéc, Ion. oikéc, roc, T6,. that which is like, esp. like truth, likely, probable, reasonable, a likelihood, Kur., etc., in Hat. usu. ra olkéra, likelihoods, 1 155, etc., 70 od. eixdc, Thuc. 2, 89: KaTa TO €iKOc, in all likelihood, Thuc. 1, 121; also 7@ eixéru, Thuc. 6, 18: mayr7i T@ olx6TL, Hdt. 3, 103: eixéru, sub. éori, it is likely, c. inf., Eur., Thuc., etc.—2. eixota, propositions generally true, likelihoods, Arist. Org. —II. reasonable, fair, equitable, Thuc. 2, 74, etc.; mapa TO eikO¢, unreasonably, Id. 2, 62: cf. émtecxgje. Neut. part. from eixa, focxa. Compar. eixd- TEPOV. EixocdBotoc, ov, poet. éeux. (elxo-ot, Bovc) worth twenty oxen, Od Eixocdedpoc, ov, (elxoot, dpa) of twenty sides or surfaces, Plat. Hixooastiec, &c, (etxoct, roc) of twenty years, Hdt. 1,136. Hence Eixocaeria, ac, 7, 4 period of twenty years, Phil. Hixooaeric, idoc, 7, pecul. fem. of elkooaeThc, a woman twenty years old, Plat. Rep. 360 E. Kixoodxce, poet. éerk. (etkoor) twen- ty times, Il. ElxoodkAtvoc, ov,=eixooikd.woc. Eixocdkwdos, ov, (eixoat, K@AoY) of twenty clauses. Eixoodkwres, av, (eixoot, KOT) with twenty oars. : Elxocdunvoe, ov, (etkoot, uqv) of twenty months, or so old, Anth. Elxocdrnyxve, v.=sikocin., kiovec, Chares ap. Ath. 538 D. EixocarAdotoc, a, ov, and Eixocariticiwy, ov, Plut. (etxocr) twenty-fold. ; Eixocde, é0oc, 7),=eixde, rare form, Sext. Emp. ' Eixocacréduoe, ov, (eikoot, oT adt- 0) of twenty stadia, Strab. tElxécatoc, ov, 6, the twentieth, Tzetz. Elxoodgvaaroc, ov, (etxoat, d0AAov) with twenty. leaves, Addor, Theophr. tElxocernpic, ioc, 7, (eixect, ét0¢) a period of twenty years, Dio C. Elxooérne, 6, fem. -étic, Loe, 7}s= elxooastync, Anth. Elxoonpne, ec, (etkoot, Gpw) with twenty banks of oars, Ath., like rpz#pne. EVKOSI, poet. éeécooz, and before a vowel éeéxoowv, of, al, td, indecl. twenty, Hom.: Dor. elxati, Sanscr. vingati, Lat. viginti. ElxociBoroc, ov,=elxoodBotac. Elxoo.va, or -d¥o, (etxoot, dtw) two and twenty. Elxootedpoc, ov,=elkoodedpoc ; el- Koolevvéa, nine and twenty, Ath.; ei- Kooves, six and twenty ; and eixootér- Ta, seven and twenty, Hipp., are all suspected by Dind., who prefers eiio- olvevvéd, etc. Hixooverie, é¢, fem. -etic, dog, ; =elxooaert7c, Dio C, Elxootxarrétpatog, 7, ov, (etkoct, Kai, Tétparoc) the twenty-fourth, Anth. ElkocixAtvoc, ov, (elkoot, KAivn) with twenty couches or seats at table, Diod. ; Eixéouuvoc, ov, (etkoot, uva) (or rather -uvéwc, Lob. Phryn. 554) of twenty minae, Lys. ap. Poll. 9, 57. Eikoow7#piroe, ov, only Il. 22, 349, elk. Growa, @ twenty-fold ransom. (From eixoou and vApitoc, vAptotoc, twenty-fold without dispute ; others from elxoot épivovTa, i. e. &toovpueva.) EIKG tElxécroz, ol, late torm fur eixuse, Anth. append, 262. 4 Elxoovoxté, (eixoot, OKO) twenty: eight, Diod. Euxoovte, ioc, 7, (elxoce-wévrTe, ETo¢) five and twenty years old, Anth. ; Elxoourévte, (eixool, TévTe) twenty-five, ap. Dem. 926, 4. Elkooinnyve, v, (etxoot, mayue) of twenty cubits, Hdt. 3, 60. Eikoo.récoapec, neut. a, (elxooTpec) twenty-four, Diod. TEikooitpeic, neut. -tpia, (eixoct) Tpeic) twenty-three, Ath. 585 B., Eixocépyviog, ov, (eikoor, dépyvid) of twenty fathoms, Xen. Cyn. 2, 5. Eixcopoc, ov, poet. éerx., with twenty oars, Od. 9, 322, cf. mevTyKéy-Topoc. Eixoar#, aia, aiov, (eikooréc) on the twentieth day, Hipp. Eixoar#, fic, 7, V. sub eixoorée I EixoaroéBdouos, ov, (eikooréc, 8 dowoc) the twenty-seventh, Plut.2,1027E, EixoAdyoc, 6,7, (eixoaTH. 15% 3} one who collects the twentieth, a.ax O} to collector, Ar. Ran. 363. Eixoore, 7, 6 v, poet. écrxoord¢ the twentieth, Hom.—lUl. 7 eixoory, tax of a twentieth, Lat. vicesima, esp one levied by the Athenians on the imports and exports of the subject allies in lieu of tribute, ‘eix. Tov yey- vouévov, TOV Kata GdAaccav, Thuc. 6, 54; 7, 28, v. Béckh P. E. 2, 38. sq- tEixooréruproc, n, ov, (eixooré¢e Tétaptoc) the twenty-fourth, Plut. Eixorévye, ov, 6, (eixoar#, Ovéo Lat) a farmer of the eixoory, like ef KooToAoyoc, Arr. EixotoAoyéa, @, (eikdéc,.Aéye) te infer from probabilities, guess, Strab Hence EixoroAoyia, ac, 7, a probability, or an inference from one, Archyt. ap. Stob. Ecl. 1, 724. Eixétwe, Att. adv. part. perf. from forxa, eika, in all likelihood, probably, as may be expected, naturally : fairly, reasonably, Aesch. Supp. 403, anc freq. in Thuc.: eixétwe¢ &yet, "tis reasonable, Eur. J. T. 911, cf. Or. 737: oft. followed by yap, Wolf. Dem. Lept. p. 252. Eixréov, verb. Adv. from exw, one must yield, Philo. TEXIX, V. sq. Eixtov, 3 dual perf., éxro, 3 dual plqpf., éxro, 3 sing. plqpf. c. pass. Signs from six, etc., to be readily yielded, pliable, Themist. ELEQ, a pres. which appears in imperf. eixe, it appeared, seemed good, only in Il. 18, 520; for its deriv. ten ses v. sub govxa. EPKQ, f. -S0, to yield, give way, draw back, retire, Hom., more strongly éricow eixery, Il. 5, 606; c. dat. pers. et gen. loci, und’ elxete yap-bung ’Apyeiotc, shrink not from the fight for them, Il. 4, 509, cf. 5, 348, elke Tpoldpov, retire from the door, Od. 18, 10, so etx. revi rH¢ 6000, Hat. 2, 80: c. dat. pers. et inf., Od. 5, 332; also absol., esp. of retreating, making way, rising from one’s seat out of respect, Il. 24, 100, Od. 2, 14: later also with éx: hence metaph. c. gen. elkety Ovjod, to withdraw from passion, give it up, Soph. Ant. 714, although Herm. reads ἀρχού, with Ald., in next sense—ll. to submit to, obey, follow, very freely in Hom, c. dat.,e. g. dug ovvy, dppadiarg, aidot elkerr, to give way to, yield to passion, folly, s'oth, sense of shame, and free. in Tag.: hence also of any impulse, διαλία, following his own bent, Il. 9, 593: 80 TH jAtKin eix. Hdt. 7, 18: also Bin” kal Kdptet Elxecy, to give one’s self up to one’s might and strength, trust therein, Od. 13, 143, revin elxwy, biased, impelled by poverty, Od. 14, 157. As this implies a state of subjection, hence—Ill. to be under, be weaker or inferior, tivé Ti, to another in a thing, Il. 22, 459, Od. 11, 515: also c. dat. rei, elxevy mddecat, to be less swift of foot, Od. 14, 221: hence in genl. to be conquered, excelled by, rive.—lYV. transit. to yield up, abandon, resign, Tivi TL, ll. 23, 337: in genl. to give, grant, allow, Lat. concedere, sAodv Tivt, Soph. Phil. 465; so too Id. O. C. 172, Plat. Legg. 781 A. (Elka of. has the digamma in Hom., so that it is well compared to Germ. weichen, Anglo-Sax. vican, and prob. to our weak.) Eixéy, %, gen. dvoc, acc. 6va, etc.; also poet. and Ion. gen. eixotc, acc. eix@, acc. pl. eixovc, but with no nom. eix@ in use, Valck. Phoen. 457, (Zo1-Ka) a figure, image, likeness, of a picture or statue, Hdt. 2, 130, 143, etc.: of needlework, Eur. I. T. 223.—I. anything like, a similitude, semblance, phantom, Eur. H. F. 1002.—2. a simile, Ar. Nub. 559, and Plat., cf Arist. Rhet. 3, 4—IL. eixéva, as adv. after the manner of, like, Lat. instar, desponativ eixova, Plat. Crat. 400 C. Hike, part. of Zovxa, q. v. Eiia, eiAduny, late 1 aor. act. and md. of aipéw formed from 2 aor., v. Buttm. Catal. p. 9. EiAadév, adv. (etAn)=tAnddv, Hat. 172. EiAadév, ov, alsowr. IAaioc, name of a month among the Delphians, Tocer. HWAdrivdla, f. -daw, (elAarivn) to feast, revel, esp. in a large company, to be a guest, Od. 2, 57; 17, 536, and Pind. Hence EiAdrivacrie, ob, 6, aafeaster, guest, booon-companion, II. 17, 577. Eldexivn, ne, 7, afeast, given by a single host, Hom., who distinguishes it from ydayoc and épavoc, but comprehends all three in dac, Od. 1, 226. (Usu. deriv. from wivewy Kar’ eiiag: acc. to others from AdmTw.) EiAap, apos, 76, (eiAw) orig. a covering, wrapping round: hence a protection, defence, vyOv Te Kai avTov, a Shelter for ship and crew, Il. 7, 338, etc.: also a fence, defence against a thing, kuaroc, Od. 5, 257, cf. Buttm. Lexil. v. eiAeiv 9. Eidapyéw, 6, to command a squadron of horse, Theb. word in Inscr., v. Müller Orchom. 470, sq.: from EtAdpyne, ov, 6, (etAn, apyw) a leader, commander of a troop or squadron of horse, esp. at Thebes; cf. iA. EiAaridne,=’Edarione, Pind. Biddrivoc, 7, ov, poet. for éAdrivoc, of fir or pine, Hom. EiAeyuac for 2éAeyua:, perf. pass. from Aéyo. Eideibvuia, ac, 7, Ilithyia, the goddesses of child-birth, who comes to aid those who are bringing forth: Hom. mentions more than one, and calls them daughters of Hera (Juno) in Il. 11, 270; 19, 119; Hes. Th. 922 speaks of one, daughter of Zeus (Jupiter) and Hera (Juno); in Pind. also 'HAei-Syca and' EAevdo, in Anth. HiApGuca, Argiyy EdAsovia j=the Roman Luci-ra; later made identical with Diana. Botigers Iithyia, Weirn. 1799. EIAl A quasi-participial form, ef. dyuva, dprvuca, from éAetoecbar, éAnAv0oévat.) Hence EiAedviag roAcc, 7, Ilithyiopolis, a city of Aegypt, Diod. S., Strab. Eitébviov, ov, 76, (ElAeiOvia) a temple of Ilithyia. EiAsde, od, 6, (elAéw) a grievous disease of the intestines, Lat. Neus volvulus, Hipp., and Aretae—II. a lurking place, den of animals, v. elAvog.—ill. a table or block used in slaughtering, a dresser, V. éhedc. EiAéacov, ov, 76, Lesium, a city of Boeotia, Il. 2, 499. Eitéw, Att. eiAéw, lengthd. form from eiAw, q. Vv. Eitéa, (€iAn) to sun. EiAeadne, e¢, (eiAéw) eidoc Eitéa, (eidoc) to sun. EiAn, 1S) patag. EiAn, 1S) patag. EiAn, ne; 7, the sun’s warmth, Ar. Vesp. 772; warmth in gen.; v. en, ahéa. EiAydov and siAnda, adv. (etAn)= iAndov.—il. (eiAéw) by rolling along, Anth. ElAnbeoéw, G,=eiAéw, to sun, baskin the sun, Hipp. : from EidAnbepie, &¢, (eiAn, 0épw) warmed by the sun, warm, Hipp. : from Epyowat: hence eiA7jAoub-ev, 1 plur. perf. Ep. for éAnAtayev, Hom. EiAnua, atoc, 76, (eiAéw) a veil, covering, wrapper, Lat. mvolucrum, ap. Stob. p. 197, 55. Il—eiAedc 1, Hipp. —Ill, late, a vault. Hence EiAnuarixoc, 4, 6v, vaulted, groined, arched. EiAnuapat, for AéAnupaz, perf. pass. of Aau Pave. Eianate, ewe, 7, Att. eid, (eiAéw) a winding, rolling: a whirl-wind. EiAnate, ewe, 7 (elAéw) a warming, sunning, Lat. apricatio: in gen). warmth, heat, Plat, Rep. 380 E. Eidnriée, 7, ov, Att. eiA rolling one’s self or others, (Qa, wrigglng animals, Arist. H. A.: from Eihyzoc, 7, 6v, Att. cid. (eiAéw) wound, twisted, twined.—II. vaulted, arched; late. EiAnoa, for A€Anoa, perf. act. of apuBava. EiAnya, for AéAnya, perf. act. of hayxavw. Kiktyyidw and etAryyoc, 6, later forms of iAvyy. EiAryua, atoc, 76, —u6¢, ov, 6, adj. —yaTtoon¢, €¢, etc., poet. and Jon, for Ly. tHiacyuar, perf. pass. from éA/saw. Eidixe, eooa, ev, and elAtcoes- Onc, €¢,=EAIK. Eidixe, eooa, ev, and elAtcoes- Onc, €¢,=EAIK. Eidixe, eooa, (stan, xpivw) examined by the Sun's light, tested, found genuine; hence—Jl. unmixed, Plat. Symp. 211 E; distinct, separate, piAa Xen. Cyr. 8, 5, 14.—2. pure, clear, un-corrupted, Hipp., Plat. ete.—3. perfect, entire, Plat. Ax. 370 C.—4. distinct, palpable, sheer; adexia, Xen. Mem. 2, 2, 3. Adv. —vide, of itself, absolutely. Plat. Rep. 477 A. The form of Arptvéu, et Arptvéu, etc., is more rare, though etymology is for it, and the best MSS. of Plato usu. have it. [cf. évévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévévév Eiduovia, ac, 7, Arg. for EiAeiOuia EdAurrodng,-ov; 6, later form for sq Ejiirove, 6,4, rovv, 76, gen. rodoc (eitAw, move) trailing-footed, esp. trawl: ing the hinder feet heavily along in walking, in Hom. (orly in dat. and acc plur.) always epith. of oxen, which trail along and plait their hind-legs as they go, v. Hipp. 785 C: absol. or owen, kine, Theocr. 25, 131; Eupol Col. 5, also uses it of women. EiAtokoiro, ewc, 9, (etAn, oxo TOW) a blind-dizziness, Lat. vertigo, elsewh. exorodivia. TEiAvcode, od, 6, Lissus, masc. p n., Qu. Sm. 1, 228. Eitiooa, poet. and Ion: for éAicow, Tl. 12, 49. Elhitevijc, éc, epith. of the plant dypwortic, Vheocr. 13, 42, acc. to some from éAo¢ and Telvw, stretching or spreading through marshes; others from eiAw and reiveo, spreading by twists and tendrils, like ivy; the first more prob, as dypworte is 2 kind of couch-grass. EiAiyaro, Jon. 3 pl. plqpf. pass. from éAicou, for eiA Tévévov, | aor. act. of Akw from a theme *éAKvo. TÉvévov, | aor. pass. of Axw, v. foreg. Bees ElAAde, 7,=lAdw, idAdc, ub. TÉvévov, less usu. 1 aor. act. from &A«Kw than eiAKvoa. EfAov and eiAéuny, aor. 2 crv. and mid. of aipéw, Hom. Eiéredov, ov, 76, rarer form for Gethéredov. EiAoya, Att. perf. from Aéyw. EiAvdwée, od, 0, (elAvw) a lurking place, den, Nic. EiAdua, aroc, 76, (elAvw) a cover wrapper, dress, clothing, Od. 6, 179 and Ap. Rh. TÉvévov, | aor. pass. from eiAdo. EiAvotc, 0d, 6,=cidvOucc, a lurking place, den, Xen. Cyn. 5, 16. Eide, boc, 7,=lAvc, mire, a morass. [o Valck. Ad. p. 248.] TÉvévoc, | aor. pass. part. from eldbo. EiAvotc, ewe, 7, Att. etfA., (elAtw) =elAnowc Also a creeping. Eidvordouat, = avor., for which it is almost always av., to wriggle along, crawl like a worm. Hence EiAvoropa, arog, TO, a worm-like wriggling motion. Eihiooo, = eidtvw, to roll along, wind, whirl, Tl. 20, 492.—II. intr. te roll one’s self forth or along, to whirl about, of blazing torch-light, Hes. Sc. EiAvodo, 6,—foreg., Il. 11, 156. EVAY’Q, Att. eiAvo, f. -dow : pert pass. elAdwar: aor. pass. part. eldvo Geic. To wind or wrap a person or thing rownd, enfold, enwrap, cover, very rare in act., as only once in Hom. viz. Il. 21, 319, wédde uty abtov et Qbo0 Waud0orcr (and this might be referred to catetAdw). Pass. to wrap one’s self rownd or about, be concealed or covered, Hom. esp. in part. perf. ei- Avpévoc as eld. Paudbo, buried; also eld. odkeol, ZaAK@, covered with shields, brass; eA. vuKti, vepé?n, veiled, shrouded in night, cloud: alsc ELAQ practi Kal Koviare eiAvTo, Il. 16, 640, ef, Od. 5, 403 ELAQ Also pass. to wind, wiggle, creep, or crawl along, Soph. Phil. 289, 701, Metagen. Thur. 1, 4. Akin to ei2u, eiAéw, etc.: Buttm. exil. in voc., assumes that eiAiw had orig. only the sign of wrapping, enfolding, éAtw, that of twisting together or winding, which agrees with fomer’s use: but later they were confused: v. efAw, fin.) [0 in Hom., except in 3 pl. perf. pass. e/Avatas: in pres., which is not found in Hom., ob, 0 in Soph.: Hin Metag. 1. c., cf. Jac. A. P. p, 588.] ; El’ AQ, also efAAw, and sometimes tAAw (q.v:). more freq. efAév, Att. elAéw, esp. in act., and in Theocr.: fut. elAjow: aor. 1, 3 plur. éAcav, inf. Acar, Ep. éAcat, part. éAcac, Hom.: perf. pass. &eAwaz, Il.: aor. Pass. édAnyv, IL, inf. GAjvat, dAnjpevat, Il., part. ddeic, eioa, év, Hom. ; in prose, also aor. 1 eiA“Onv, but prob. only in compos., cf. KatetAéo : plqpf. 3 sing. 6AnTo, Ap. Rh.: an aor. 1 mid. 7Acauny, and a still more strange aor. 2 7Aedunv, are quoted only from Simon, and Ibyc., v. subvoce. Radical. Sign of act. to roll or twist tight up, hence to press hard or close, e.g. of a warrior who presses the enemy close, Il. 8, 215; Aadv Kara rel(yea EAoat, to force the host back to the walls, Il. 21, 295, cf. 225; so Kara Tpbuvac or éxt rptuvyoryr, Il. 1, 409, etc.: to force together: hence to coop, block up, shut up in a thing, crowd together, évi oni, évi oni, év oreiver, Od. 12, 210; 22, 460; c. dat. only, Il. 18, 294: metaph. of a storm, which drives a ship along or about, Il. 2, 294, Od. 19, 200: via kepavv@ éAcac, striking the ship with a thunderbolt, Od. 5, 132; 7, 250. In act. Hom. has only e/Aéa, never ciAw.—ll pass. and mid. to crowd, be rolled all up together, to throng together, Il. 5, 782: to be shut, cooped up or in, of the besieged, Il. 5, 203, Evi vvvoi, Il. 12, 38, é¢ worapov el- Aevvto, they were pushed into the river, Il. 21, 8: metaph., Avd¢ Bov- Ajowy éseApévoc, straitened, held in check by the counsels of Jupiter, Il. 13, 524: to throng together, assemble, crowd thackly together, dpi Avoundea eiAd- ue vot, Il. 5, 782: this sign is very free in Homer, in aor. pass. éAnv, esp. of a routed army; which however does not justify us in inferring a sign to retreat, recoil, as some have done in Il. 5, 823, etc.: drev bdwp, water collected, ponded, Il. 23, 420; also to draw one’s self together, crouch, cower, im’ donidt, Il. 13, 508; 20, 278; also ’AyiAja dAele péver, collecting himself he ated the attack of Achilles, Il. 21, 571; so of a lion which gathers itself for a bound, Il.
|
https://got.wikipedia.org/wiki/%F0%90%8D%86%F0%90%8C%B0%F0%90%8C%B9%F0%90%8D%82%F0%90%8C%BD%F0%90%8C%B9%F0%90%8C%B6%F0%90%8D%89%20%F0%90%8D%86%F0%90%8C%BF%F0%90%8C%B8%F0%90%8C%B0%F0%90%8D%82%F0%90%8C%BA
|
Wikipedia
|
Open Web
|
Wikimedia/Pleias
|
CC-By-SA
| 2,023
|
𐍆𐌰𐌹𐍂𐌽𐌹𐌶𐍉 𐍆𐌿𐌸𐌰𐍂𐌺
|
https://got.wikipedia.org/w/index.php?title=𐍆𐌰𐌹𐍂𐌽𐌹𐌶𐍉 𐍆𐌿𐌸𐌰𐍂𐌺&action=history
|
Gothic
|
Written
| 22
| 561
|
𐍆𐌰𐌹𐍂𐌽𐌴𐌹𐍃 𐍆𐌿𐌸𐌰𐍂𐌺 𐌹𐍃𐍄 𐍃𐍉 𐍆𐌰𐌹𐍂𐌽𐌾𐍉 𐌱𐍉𐌺𐌰𐍄𐌴𐍅𐌰 𐍂𐌿𐌽𐍉. 𐌱𐍂𐌿𐌺𐌽𐌰𐌳𐌰 𐍅𐌰𐍂𐌸 𐌰𐌽𐌰 𐌲𐌰𐌹𐍂𐌼𐌰𐌽𐌹𐍃𐌺𐌰𐌹𐌼 𐌺𐌿𐌽𐌾𐌰𐌼 𐌹𐌽 𐌼𐌹𐌳𐌿𐌼𐌰𐌹 𐌰𐌷𐍄𐌿𐌳𐌹𐌽𐍃 𐌾𐌴𐍂𐌰𐌷𐌿𐌽𐌳𐌹𐍃.
𐍆𐌿𐌸𐌰𐍂𐌺
𐌲𐌰𐍅𐌹𐍃𐍃
Runenprojekt (𐌸𐌹𐌿𐌳𐌹𐍃𐌺𐌰 𐍂𐌰𐌶𐌳𐌰)
𐌱𐍉𐌺𐌰𐍄𐌴𐍅𐌰
|
（2021）辽0726民初1958号
|
Chinese-Court-Decisions
|
Open Government
|
Pleias
|
Public Domain
| 2,021
|
黑山富洪物业有限公司与李明物业服务合同纠纷一审民事裁定书
|
辽宁省黑山县人民法院
|
Chinese
|
Written
| 357
| 278
|
辽宁省黑山县人民法院 民 事 裁 定 书 （2021）辽0726民初1958号 原告：黑山富洪物业有限公司，住所地黑山县黑山镇一街中大中路金鼎御锦城二期A座门市由北向南向东第24户。 法定代表人：许兰英。 被告：李明，男，26岁，汉族，住黑山县。 原告黑山富洪物业有限公司与被告李明物业服务合同纠纷一案，在本院审理中，原告黑山富洪物业有限公司于2021年7月5日向本院提出撤诉申请。 本院认为，当事人有权在法律规定的范围内处分自己的诉讼权利，现原告要求撤回起诉，符合法律规定，本院予以准许。本院依照《中华人民共和国民事诉讼法》第一百四十五条一款之规定，裁定如下： 准许原告撤诉。 案件受理费25元，由原告负担。 审判员 曹磊 二〇二一年七月五日 法官助理卓慧 书记员梁怡 来自：www.macrodatas.cn
|
（2019）黔2301执594号
|
Chinese-Court-Decisions
|
Open Government
|
Pleias
|
Public Domain
| 2,019
|
贺正虎、周洪秀与蒋加芹失信决定书
|
兴义市人民法院
|
Chinese
|
Written
| 1,103
| 762
|
兴义市人民法院 失 信 决 定 书 （2019）黔2301执594号 本院在执行申请执行人贺正虎、周洪秀与被执行人蒋加芹一案中，经查，被执行人具有《最高人民法院关于公布失信被执行人名单信息的若干规定》第一条第一项规定的情形。依照《中华人民共和国民事诉讼法》第二百五十五条、《最高人民法院关于公布失信被执行人名单信息的若干规定》第一条第一项的规定，决定如下： 将蒋加芹纳入失信被执行人名单。 本院将根据《最高人民法院关于公布失信被执行人名单信息的若干规定》的规定，将失信被执行人名单信息录入全国法院失信被执行人名单库，并向社会公布；同时将失信被执行人名单信息向政府相关部门、金融监管机构、金融机构、承担行政职能的事业单位及行业协会等通报，供相关单位依照法律、法规和有关规定，在政府采购、招标投标、行政审批、政府扶持、融资信贷、市场准入、资质认定等方面，对失信被执行人予以信用惩戒；将失信被执行人名单信息向征信机构通报，并由征信机构在其征信系统中记录；国家工作人员、人大代表、政协委员等被纳入失信被执行人名单的，将失信情况通报其所在单位和相关部门；国家机关、事业单位、国有企业等被纳入失信被执行人名单的，将失信情况通报其上级单位、主管部门或者履行出资人职责的机构。 本院将根据《最高人民法院关于限制被执行人高消费及有关消费的若干规定》第一条的规定，对纳入失信被执行人名单的被执行人采取限制消费措施。被执行人为自然人的,被采取限制消费措施后，不得有以下高消费及非生活和工作必需的消费行为：（一）乘坐交通工具时，选择飞机、列车软卧、轮船二等以上舱位；（二）在星级以上宾馆、酒店、夜总会、高尔夫球场等场所进行高消费；（三）购买不动产或者新建、扩建、高档装修房屋；（四）租赁高档写字楼、宾馆、公寓等场所办公；（五）购买非经营必需车辆；（六）旅游、度假；（七）子女就读高收费私立学校；（八）支付高额保费购买保险理财产品；（九）乘坐G字头动车组列车全部座位、其他动车组列车一等以上座位等其他非生活和工作必需的消费行为。被执行人为单位的，被采取限制消费措施后，被执行人及其法定代表人、主要负责人、影响债务履行的直接责任人员、实际控制人不得实施前述行为。因私消费以个人财产实施前述行为的，可以向本院提出申请。被限制消费的被执行人因生活或者经营必需而进行前述禁止的消费活动的，应当向本院提出申请，获批准后方可进行。 如违反规定进行消费，经查证属实的，本院将依照《中华人民共和国民事诉讼法》第一百一十一条的规定，予以罚款、拘留；情节严重，构成犯罪的，依法追究刑事责任。 本决定一经作出即生效。 二〇一九年八月十九日 来自马克数据网
|
（2020）鲁0783民初3283号
|
Chinese-Court-Decisions
|
Open Government
|
Pleias
|
Public Domain
| 2,020
|
寿光市人民政府古城街道办事处与刘传福、周香芳房屋买卖合同纠纷一审民事裁定书
|
山东省寿光市人民法院
|
Chinese
|
Written
| 395
| 319
|
文书内容山东省寿光市人民法院民 事 裁 定 书（2020）鲁0783民初3283号原告:寿光市人民政府古城街道办事处。住所地：寿光市古城街道。统一社会信用代码：11370783004315712K。法定代表人：尹爱军，主任。被告：刘传福，男，1965年12月19日生，汉族，住寿光市。被告：周香芳，女，1965年06月21日生，汉族，住寿光市。原告寿光市人民政府古城街道办事处与被告刘传福、周香芳房屋买卖合同纠纷一案，本院于2020年6月17日立案。原告寿光市人民政府古城街道办事处未在本院指定期限内预交案件受理费。依照《中华人民共和国民事诉讼法》第一百一十八条、第一百五十四条第一款第十一项、《最高人民法院关于适用的解释》第二百一十三条规定，裁定如下：本案按寿光市人民政府古城街道办事处撤回起诉处理。审判员 张大鹏二〇二〇年六月二十九日书记员 刘柄浩 关注公众号“马 克 数 据 网”
|
https://uk.wikipedia.org/wiki/%D0%9B%D1%83%D1%86%D1%96%D0%B9%20%D0%95%D0%BC%D1%96%D0%BB%D1%96%D0%B9%20%D0%9F%D0%B0%D0%B2%D0%BB%D0%BE%20%D0%9C%D0%B0%D0%BA%D0%B5%D0%B4%D0%BE%D0%BD%D1%81%D1%8C%D0%BA%D0%B8%D0%B9
|
Wikipedia
|
Open Web
|
Wikimedia/Pleias
|
CC-By-SA
| 2,023
|
Луцій Емілій Павло Македонський
|
https://uk.wikipedia.org/w/index.php?title=Луцій Емілій Павло Македонський&action=history
|
Ukrainian
|
Written
| 1,671
| 5,060
|
Луцій Емілій Павло Македонський (; близько 229 до н. е. — ) — політичний, державний і військовий діяч Стародавнього Риму, двічі консул , видатний воєначальник.
Родина
Луцій Емілій Павло Македонський народився близько 229 до н. е. Його батьком був Луцій Емілій Павло, консул, якого було вбито у битві при Каннах у 216 до н. е. Луцій Емілій походив з гілки Еміліїв Павлів давнього патриціанського роду Еміліїв. Вплив цього роду був неосяжним, зокрема завдяки його вдачі та союзу з Корнеліями Сципіонами. Луцій Емілій Павло Македонський був батьком Сципіона Еміліана Африканського і братом Емілії Терції, дружини Сципіона Африканського
Рання кар'єра
До 195 до н. е. його тричі було обрано військовим трибуном, був квестором.
У 194 до н. е. призначений одним з триумвірів для заснування колонії в Кротоні. Після завершення військової служби, Емілія Павла у 193 до н. е. його обрано curule aedile — курульним еділом Він виграв вибори у 12 суперників. На цій посаді, разом з колегою Марком Лепідом, засудив низку скотопромисловців за зловживання суспільними землями і на стягнені кошти прикрасив храм Юпітера, побудував два портика. Близько 192 до н. е. увійшов до колегії авгурів.
Наступним кроком у його cursus honorum було обрання претором у 191 до н. е. Під час перебування на цій посаді Павло вирушив до провінції Іспанія, де у 190-189 до н. е. отримав проконсульські повноваження.
Упродовж 191-189 до н. е. воював проти лузитанів, спочатку зазнав поразки у битві при Ліконі, але потім одержав вирішальну перемогу і відновив порядок у провінції, проявивши при цьому справедливість та безкорисливість.
У 189 до н. е. увійшов до складу комісії з десяти легатів для розгляду азійських справ і укладання мирного договору з царем Антіохом. Ця комісія протистояла проконсулу Гнею Манлію Вульсону, який провокував конфлікт з Антіохом і самочинно розв'язав війну проти галатів. У 187 до н. е. від імені комісії виступив порти надання йому тріумфу.
Повернувшись до Риму, Емілій Павло почав добиватися консульства, але тричі програвав вибори. Розповідаючи про вибори 185 року до н. е., Тит Лівій називав Павла «досвідченим здобувачем, який вважав, що має тим більше право на консульство, що один раз йому вже було в ньому відмовлено» .
Вперше обрано консулом у 182 до н. е.; другим, молодшим консулом став Гней Бебій Тамфіл. Обидва консули відправилися до Лігурії, де вдало вели воєнні дії.
У 181 до н. е. їхні повноваження були продовжені, й Емілій Павло вирушив у похід проти лігурійців-інгавнів. Противник осадив його табір, але римляни зробили несподівану вилазку і здобули перемогу, після якої все плем'я здалося. Після повернення до Риму отримав тріумф.
Згідно Плутарху, Луцій Емілій «багаторазово виявляв недвозначне бажання знову отримати посаду консула» і один раз висунув свою кандидатуру, але програв вибори. Після цього він відмовився від участі в політиці, зосередившись на своїх жрецьких обов'язках і вихованні синів.
У 171 до н. е. входив до складу комісії з розслідування здирства римських намісників у Дальній Іспанії, але, з чуток, сам перешкоджав засудженню звинувачуваних.
Луцій Емілій Павло та Македонія
Третя Македонська війна
Третя Македонська війна розпочалася у 171 до н. е., коли македонський цар Персей розбив римську армію під керівництвом Публія Ліцинія Красса у битві при Калліцині. Протягом двох років війна точилася з перемінним успіхом для обох сторін, аж поки у 168 до н. е. консулом не було вдруге обрано Луція Емілія Павла. Як консула його призначив сенат для вирішення македонського питання. Здійснивши усі необхідні приготування до війни, навесні Емілій Павло відплив до Македонії до римської армії, яка перебувала біля Геракліона. За допомогою обхідного маневру він змусив македонське військо відступити до Підни і там 22 червня 168 до н. е. здобув переконливу перемогу. Напередодні битви відбулося місячне затемнення, що злякало воїнів в обох арміях. Згідно Цицерону, римлян заспокоїв, пояснивши їм суть явища, військовий трибун Гай Сульпиций Галл; за іншими джерелами, він навіть передбачив затемнення і все пояснив війську заздалегідь. У македонців була перевага в чисельності: 40 000 піхотинців і 4 тисячі вершників проти 26 000 римлян. Проте Персей довго не починав битву, розраховуючи, що атакувальною стороною стануть римляни. Луцій Емілій зі свого боку теж вичікував. Нарешті, ввечері македонська фаланга рушила в бій і добилася великих успіхів. Битва вже зав'язалася, коли з'явився Емілій і побачив, що македонці в перших лініях встигли встромити вістря своїх саріссамі в щити римлян і, таким чином, стали недосяжні для їхніх мечів. Коли ж і всі інші македонці за обумовленим сигналом разом відвели щити від плеча і, взявши списа напереваги, стійко зустріли натиск римлян, йому стала зрозуміла вся сила цього зімкнутого ладу, що грізно наїжився; ніколи в житті не бачив він нічого страшнішого і тому відчув переляк і замішання, і нерідко згодом згадував про це видовище і про враження, яке воно залишило. — Плутарх, Емілій Павло
Римляни, хоча і не почали тікати, були витіснені до свого табору і до гори Олокр. Але незабаром ситуація почала змінюватися: бій швидко перетворилося на бійню, до якої приєднався римський флот. Сильна македонська кіннота так і не взяла участь в битві, що зробило перемогу римлян ще повнішою. 20 тисяч македонців були вбиті, ще 11 тисяч потрапили в полон. Цар Персей втік до Самофракії, але незабаром здався римлян у полон і був доставлений до Емілія Павла. Третю Македонську війну було завершено.
Пізніші роки
У 167 до н. е. залишився у Македонії на посаді проконсула; здійснив подорож до Греції. Разом з комісією децемвирів врегулював справи в Македонії та Греції, задовольнив скарги грецьких держав-союзників, та, аби показати приклад, наказав вбити 500 македонців, відомих своєю прихильністю царю Персею. Також він відправив багато людей у вигнання до Італії та конфіскував їх скарб в ім'я Риму, але згідно з Плутархом, забрав значну частину трофеїв собі.
При поверненні до Риму в 167 до н. е., солдати були незадоволені своєю долею награбованого, тому, щоб задовольнити їх, Емілій Павло вирішив зробити зупинку в Епірі, країні, яка підозрювалася у симпатіях до Македонії. Хоча в Епірі вже було відновлено мир, Павло, згідно з наказом сенату, віддав на пограбування 70 його міст; 150 000 людей були захоплені в рабство і країна залишилася спустошеною.
Повернення Павла до Риму було розкішним. Захопивши нечувану здобич у Македонії та Епірі й полонивши самого македонського царя, Емілій Павло мав ефектний тріумф, попри те, що його солдати були незадоволені рішенням командувача повністю передати казну Персея державі й намагалися завадити призначенню його тріумфу. Проте тріумф все ж таки відбувся 28 листопада-30 листопада 167 до н. е. Здобич у македонській війні була настільки великою, що в Римі з того часу припинилося стягнення трибуту. Як жест визнання, Сенат нагородив Луція Емілія Павла агноменом (прізвиськом) Македонський (Macedonicus). То була вершина його кар'єри.
Останні роки життя і смерть
У 164 до н. е. обіймав посаду цензора. У 162 до н. е. ймовірно був інтеррексом. Наприкінці життя Луцій Емілій Павло Македонський тяжко хворів. Хвороба то наступала, то відступала. Емілій Павло навіть провів певний час в Елеї Італійській, відновлюючи здоров'я. Згодом він повернувся до Риму і навіть приніс жертву богам. Проте хвороба несподівано ускладнилась і Емілій Павло помер у 160 р. до н. е. Про великі почесті для Емілія Павла і про любов народу до нього розповів Полібій, описуючи похорони : Самі похорони його гідні захоплення: ревне участь всіх присутніх вшанувало доблесть покійного найпрекраснішими і завидними похоронними дарами. То було не золото, не слонова кістка, не показна пишність оздоблення, але душевна схильність, повага і любов не тільки співгромадян, а й противників. Усі іспанці, лігури і македонці, скільки їх не було тоді в Римі, зібралися навколо похоронного одра, молоді та сильні підняли його на плечі й понесли, а люди старшого віку рушили слідом, називаючи Емілія благодійником і рятівником їх рідній землі. — Плутарх, Емілій Павло.
Нащадки Луція Емілія Павла Македонського
Вперше Луцій Емілій Павло одружився з Папірією Масонією, донькою консула 231 року до н. е. Гая Папірія Масона, з якою він згодом розлучився з невідомої причини. Від цього шлюбу в нього було четверо дітей: два сина та дві доньки. Старша донька, Емілія Павла Перша, ймовірно, стала дружиною сина Марка Порція Катона, а молодша, Емілія Павла Секунда, вийшла заміж за Елія Туберона, багатія з плебейської родини. Згідно з римськими істориками, Емілій Павло розлучився з дружиною, коли їхній молодший син був ще дитиною, звідси розлучення мало місце у 183-182 роках до н. е. попри це, його було обрано консулом у 182 році до н. е. тоді Павло одружився вдруге, ім'я його другої дружини невідоме. У цьому шлюбі народилося ще двоє синів, старший народився близько 181 року до н. е., а молодший — близько 176 року до н. е. Також, в Емілія Павла, ймовірно, народилася ще одна донька, Емілія Терція.
Приблизно між 175 та 170 роком до н. е. він віддав двох старших синів на всиновлення до інших родин: старшого всиновив Квінт Фабій Максим, він став Квінтом Фабієм Максимусом Еміліаном, вступаючи таким чином до роду Фабіїв. Молодшого сина, який, ймовірно, спершу носив ім'я Луцій, всиновив його ж двоюрідний брат Публій Корнелій Сципіон, старший син і спадкоємець Публія Корнелія Сципіона Африканського, і він став Публієм Корнелієм Сципіоном Еміліаном.
Після того, як старші сини були всиновлені двома найвпливовішими патриціанськими родинами Риму, Емілій Павло розраховував на те, що двоє молодших продовжать його династію, але сподівання були марними. Обидва сини померли ще у дитинстві, один за одним, в той час, коли Павло святкував свій тріумф. За Полібієм, старшому з цих дітей було 14, а молодшому 9 років; їхні імена в історії не збереглися.
По смерті Емілія Павла його сини Квінт Фабій Максим Еміліан та Публій Корнелій Сципіон Еміліан отримали його майно у спадок згідно із заповітом, хоча юридично вони вже не належали до Еміліїв Павлів. Сципіон віддав свою частку старшому братові, оскільки той був менш заможний. Друга дружина Емілія Павла (її ім'я залишилося невідомим) забрала свій посаг, продавши деяке майно чоловіка (Лівій та Полібій стверджують, що Павло помер відносно бідним, і що його здобич після македонської кампанії була невеликою).
Із смертю Луція Емілія Павла Македонського лінія Еміліїв Павлів згасла, хоча двоє нащадків і вижили. Старший, Фабій Еміліан, згодом став консулом і мав щонайменше одного сина, який в свою чергу у 121 році до н. е. став консулом Фабієм Аллобрігіком. Молодший, відомий під іменем Сципіона Еміліана, помер, не залишивши нащадків. З-поміж дочок старша була матір'ю двох консулів, а молодша донька — матір'ю консула Квінта Елія Туберона.
Див. також
Третя Македонська війна
Битва при Підні
Примітки
Джерела
Тит Лівій
Полібій
Плутарх
Давньоримські військовики
Давньоримські політики
Інтеррекси
Претори
Еділи
Емілії
|
https://github.com/fachriagustian12/sis_pakar/blob/master/application/controllers/Diagnosa.php
|
Github Open Source
|
Open Source
|
BigCode/Github/Pleias
|
LicenseRef-scancode-unknown-license-reference, MIT
| 2,021
|
sis_pakar
|
fachriagustian12
|
PHP
|
Code
| 158
| 729
|
<?php
defined('BASEPATH') or exit('No direct script access allowed');
class Diagnosa extends CI_Controller
{
public function __construct()
{
parent::__construct();
$this->isLogin = $this->session->userdata('isLogin');
if ($this->isLogin == 0) {
redirect(base_url());
}
$this->id = $this->session->userdata('id');
$this->load->model('model_diagnosa');
}
public function listDiagnosa()
{
$page['page'] = 'diagnosa';
$data['listDiagnosa'] = $this->model_diagnosa->getAll();
$this->load->view('back/template/header');
$this->load->view('back/template/sidebar');
$this->load->view('back/diagnosa', $data);
$this->load->view('back/template/footer');
}
public function getAlldiagnosa()
{
$diagnosa = $this->model_diagnosa->getAll();
echo json_encode($diagnosa);
}
public function diagnosaById()
{
$id = $this->input->post('id');
$diagnosa = $this->model_diagnosa->getById($id);
$output = array(
'kode_diagnosa' => $diagnosa->kode_diagnosa,
'nama_diagnosa' => $diagnosa->nama_diagnosa,
'keterangan' => $diagnosa->keterangan
);
echo json_encode($output);
}
public function doDiagnosa()
{
$operation = $this->input->post('operation');
if ($operation == "Tambah") {
$data = array(
'kode_diagnosa' => $this->input->post('kode_diagnosa'),
'nama_diagnosa' => $this->input->post('nama_diagnosa'),
'keterangan' => $this->input->post('keterangan')
);
$process = $this->model_diagnosa->tambah($data);
} else if ($operation == "Edit") {
$id = $this->input->post('id_diagnosa');
$data = array(
'kode_diagnosa' => $this->input->post('kode_diagnosa'),
'nama_diagnosa' => $this->input->post('nama_diagnosa'),
'keterangan' => $this->input->post('keterangan')
);
$process = $this->model_diagnosa->edit($id, $data);
}
echo json_encode($process);
}
public function deleteDiagnosa()
{
$id = $this->input->post('id');
$process = $this->model_diagnosa->delete($id);
echo json_encode($process);
}
}
|
https://www.wikidata.org/wiki/Q32533154
|
Wikidata
|
Semantic data
|
Pleias
|
CC0
| null |
Category:1986–1987 Nordic combined skiing season
|
None
|
Multilingual
|
Semantic data
| 143
| 466
|
Kategori:Nordisk kombination-säsongen 1986/1987
Wikimedia-kategori
Kategori:Nordisk kombination-säsongen 1986/1987 instans av Wikimedia-kategori
Kategori:Nordisk kombination-säsongen 1986/1987 föregås av Kategori:Nordisk kombination-säsongen 1985/1986
Kategori:Nordisk kombination-säsongen 1986/1987 följs av Kategori:Nordisk kombination-säsongen 1987/1988
Kategori:Nordisk kombination-säsongen 1986/1987 kategorin kombinerar ämnen ettårsperiod 1986-1987
Kategori:Nordisk kombination-säsongen 1986/1987 kategorin kombinerar ämnen nordisk kombination
Category:1986–1987 Nordic combined skiing season
Wikimedia category
Category:1986–1987 Nordic combined skiing season instance of Wikimedia category
Category:1986–1987 Nordic combined skiing season follows Category:1985–1986 Nordic combined skiing season
Category:1986–1987 Nordic combined skiing season followed by Category:1987–1988 Nordic combined skiing season
Category:1986–1987 Nordic combined skiing season category combines topics 1986-1987 one-year-period
Category:1986–1987 Nordic combined skiing season category combines topics Nordic combined
Kategorie:Nordische Kombination-Saison 1986/87
Wikimedia-Kategorie
Kategorie:Nordische Kombination-Saison 1986/87 ist ein(e) Wikimedia-Kategorie
Kategorie:Nordische Kombination-Saison 1986/87 Vorgänger Kategorie:Nordische Kombination-Saison 1985/86
Kategorie:Nordische Kombination-Saison 1986/87 Nachfolger Kategorie:Nordische Kombination-Saison 1987/88
Kategorie:Nordische Kombination-Saison 1986/87 Kategorie kombiniert die Themen Einjahresperiode 1986-1987
Kategorie:Nordische Kombination-Saison 1986/87 Kategorie kombiniert die Themen Nordische Kombination
|
https://www.regulations.gov/document/EPA-R09-OAR-2011-0382-0015
|
reg_docs
|
Open Government
|
kl3m
|
Public Domain
| 2,011
|
CALIFORNIA AIR RESOURCES BOARD
|
United States Government
|
English
|
Written
| 900
| 1,404
|
CALIFORNIA AIR RESOURCES BOARD
SIP COMPLETENESS CHECKLIST
(Electronic Format)
*** TO BE COMPLETED BY DISTRICT AND RETURNED TO ARB ***
All rules submitted to the EPA as State Implementation Plan (SIP) revisions must be supported by certain information
and documentation for the rule packages to be deemed complete for review by the EPA. Rules will not be evaluated for
approvability by the EPA unless the submittal packages are complete. To assist you in determining that all necessary
materials are included in rules packages sent to the ARB for submittal to the EPA, please fill out the following form and
include it with the rule package you send ARB. See the ARB's Guidelines on the Implementation of the 40 CFR 51,
Appendix V, for a more detailed explanation than is provided here. Adopted rules and rule amendments should be
checked against U.S. EPA's Guidance Document for Correcting Common VOC & Other Rule Deficiencies (Little Blue
Book, August 21, 2001) to ensure that they contain no elements which will result in disapproval by EPA.
District: Sacramento Metropolitan Air Quality Management District
Rule No: 414
Rule Title: Water Heaters, Boilers and Process Heaters Rated Less than 1,000,000 Btu per Hour
Date Adopted or Amended: Amended March 25, 2010
ADMINISTRATIVE MATERIALS
Note: All documents should be in electronic format. Items that have signatures, initials, or stamps may be scanned.
Not
Attached Attached N/A
COMPLETE COPY OF THE RULE: Provide an unmarked copy of the entire
rule as adopted or amended by your District Board.
UNDERLINE AND STRIKEOUT COPY OF THE RULE: If an amended rule,
provide a complete copy of the rule indicating in underline and strikeout
format all language which has been added, deleted, or changed since the rule
was last adopted or amended.
COMPLETE COPY OF THE REFERENCED RULE(S): For any rule which
includes language specifically referencing another rule, a copy of that other
rule must also be submitted, unless it has already been submitted to EPA as
part of a previous SIP submittal.
PUBLIC NOTICE EVIDENCE: Include a copy of the local newspaper clipping
certification(s), stating the date of publication, which must be at least 30 days
before the hearing. As an alternative, include a copy of the actual published
notice of the public hearing as it appeared in the local newspaper(s). In this
case, however, enough of the newspaper page must be included to show the
date of publication. The notice must specifically identify by title and number
each rule adopted or amended.
RESOLUTION/MINUTE ORDER: Provide the Board Clerk certified resolution
or minute order. This document must include certification that the hearing
was held in accordance with the information in the public notice. It must also
list the rules that were adopted or amended, the date of the public hearing,
and a statement of compliance with California Health and Safety Code
Sections 40725-40728 (Administrative Procedures Act).
PUBLIC COMMENTS AND RESPONSES: Submit copies of written public
comments made during the notice period and at the public hearing. Also
submit any written responses prepared by the District staff or presented to the
District Board at the public hearing. A summary of the public comments and
responses is adequate. If there were no comments made during the notice
period or at the hearing, please indicate N/A to the left.
5/1/2002
CALIFORNIA AIR RESOURCES BOARD
SIP COMPLETENESS CHECKLIST
(Electronic Format)
TECHNICAL MATERIALS
Not
Attached Attached N/A
RULE EVALUATION FORM: See instructions for completing the Rule
Evaluation Form and the accompanying sample form.
NON-EPA TEST METHODS: Attach all test methods that are referenced in
your rule that do not appear in 40 CFR 51, 60, 61, 63, or have not been
previously submitted to EPA. EPA methods used in other media such as
SW846 for solid waste are not automatically approved for air pollution
applications. Submittal of test methods that are not EPA-approved should
include the information and follow the procedure described in Region 9’s
“Test Method Review & Evaluation Process.”
MODELING SUPPORT: Provide if appropriate. In general, modeling support
is not required for VOC and NOx rules to determine their impacts on ozone
levels. Modeling is required where a rule is a relaxation that affects large
sources (> 100 TPY) in an attainment area for SO2, directly emitted PM10,
CO, or NOx (for NO2 purposes). In cases where EPA is concerned with the
impact on air quality of rule revisions which relax limits or cause a shift in
emission patterns in a nonattainment area, a reference back to the approved
SIP will be sufficient provided the approved SIP accounts for the relaxation
and provided the approved SIP used the current EPA modeling guidelines. If
current EPA modeling guidelines were not used, then new modeling may be
required.
ECONOMIC AND TECHNICAL JUSTIFICATION FOR DEVIATIONS FROM
EPA POLICIES: The District staff report or other information included with
the submittal should discuss all potential relaxations or deviations from RACT,
RACM, BACT, BACM, enforceability, attainment, RFP, or other relevant EPA
requirements. This includes, for example, demonstrating that exemptions or
emission limits less stringent than the presumptive RACT (e.g., a CTG) meet
EPA’s 5 percent policy, and demonstrating that all source categories
exempted from a RACM/BACM rule are de minimus according to EPA’s
RACM/BACM policy.
ADDITIONAL MATERIALS: Provide District staff reports and any other
supporting information concerning development of the rule or rule changes.
This information should explain the basis for all limits and thresholds
contained in the rule.
|
https://ja.wikipedia.org/wiki/%E3%81%AE%E3%81%84%E3%81%A1%E9%A7%85
|
Wikipedia
|
Open Web
|
Wikimedia/Pleias
|
CC-By-SA
| 2,023
|
のいち駅
|
https://ja.wikipedia.org/w/index.php?title=のいち駅&action=history
|
Japanese
|
Written
| 79
| 1,382
|
のいち駅（のいちえき）は、高知県香南市野市町西野にある、土佐くろしお鉄道（TKT）ごめん・なはり線の駅。香南市の代表駅である。駅番号はGN37。
歴史
2002年（平成14年）7月1日：開業。
駅構造
交換設備のある相対式2面2線のホームを持つ高架駅である。1番線を上下主本線、2番線を上下副本線とした一線スルーとなっており、設備も両方向の入線・発車が想定されているが、すべての定期列車が停車するため、ホームは方向別に使い分けられている。
駅員無配置ではあるが、駅舎内の売店で定期券や回数券を販売しており、乗車券自動販売機も設置されている。そのため、業態としては簡易委託駅に近い。入口は西入口と北入口があるが、時間帯によっては北入口は閉鎖される。バリアフリー対策エレベーター設置駅。
朝夕の混雑時を中心に、後免駅 - のいち駅間では、ワンマン扱いの列車にも車掌が乗り込むことが多い。
のりば
※上表の路線名は旅客案内上の名称（愛称）で表記している。
駅周辺
駅の開業によって、駅の西側を南北に走る道路が整備され、その道沿いに様々な店舗の出店が相次いだ。ショッピングセンターが隣接しているほか、周辺にはレストランやコンビニエンスストアなどがある。
北側
香南市役所（旧・野市町役場）
香南市立野市小学校
香南市立野市幼稚園
香南市立野市図書館
のいちふれあいセンター
野市郵便局
四国自動車博物館
南側
国道55号（南国バイパス）
南国警察署のいち駐在所
バス路線
のいち駅
野市龍河洞通
野市
駅南約100m
その他
当駅のよしかわ駅方面で高架区間が国道55号の築堤を地平区間としてアンダークロスしている箇所がある。この築堤は国道55号が土佐電気鉄道安芸線を立体交差で乗り越すために設けられたもので、ここはもともとは国道の築堤を高々架で乗り越す計画で、工事凍結の時点で一部高々架の高い構造物が出来上がっていて、これを土佐の「バベルの塔」と呼ばれていた。工事再開の際、高々架構造では建設費が高くつき、また高知東部自動車道が横断する計画があることから再利用することなく取り壊され、国道との立体交差を流用する形で国道の下を潜るよう改められ、周辺に踏切が設置された。
計画当初の仮称は漢字で「野市駅」だった。
土佐電気鉄道安芸線の時代は、野市駅と日章駅（現・立田駅）間に「西野市駅」があった。
イメージキャラクター
名称は「のいちんどんまん」。野市町で行われる「ちんどんコンクール」にちなんだキャラクターで、ピエロのような服装をしている。
なお、このキャラクターのモニュメントは北入口付近に設置されている。
隣の駅
土佐くろしお鉄道
ごめん・なはり線
快速
後免町駅 (GN39) - のいち駅 (GN37) - あかおか駅 (GN35)
普通
立田駅 (GN38) - のいち駅 (GN37) - よしかわ駅 (GN36)
脚注
参考文献
関連項目
日本の鉄道駅一覧
外部リンク
主要駅のご案内 のいち駅 - 土佐くろしお鉄道
ゴトゴトWeb - ごめん・なはり線活性化協議会
あなたの駅前物語 のいち駅（高知県） - テレビ朝日
高知県の鉄道駅
いち
土佐くろしお鉄道の鉄道駅
のいちえき
2002年開業の鉄道駅
香南市の建築物
|
End of preview. Expand
in Data Studio
Common Corpus
Common Corpus is the largest open and permissible licensed text dataset, comprising 2.27 trillion tokens (2,267,302,720,836 tokens). It is a diverse dataset, consisting of books, newspapers, scientific articles, government and legal documents, code, and more. Common Corpus has been created by Pleias in association with several partners.
Common Corpus differs from existing open datasets in that it is:
- Truly Open: contains only data that is either uncopyrighted or permissively licensed
- Traceable: each individual document is associated with documented contextual information, including licensed use or lack of copyright.
- Multilingual: mostly representing English and French data, but contains data for 8 languages with more than 10 billion tokens (German, Spanish, Italian, Polish, Greek, Latin) and 33 languages with more than 1 billion tokens.
- Diverse: consisting of scientific articles, government and legal documents, code, and cultural heritage data, including books and newspapers
- Extensively Curated: spelling and formatting has been corrected from digitized texts, harmful and toxic content has been removed, and content with low educational content has also been removed.
The dataset in its entirety meets the requirements of the Code of Conduct of the AI Act and goes further than the current requirements for data transparency. It aims to set a new standard of openness in AI, showing that detailed provenance at a granular document level is a realistic objective, even at the scale of 2.3 trillion tokens.
Common Corpus makes it possible to train model compatible with the Open Source Initiative’s definition of open-source AI, which includes openness of use, meaning use is permitted for “any purpose and without having to ask for permission". Based on the available licensing information Common Corpus can be filtered to only include public domain works or a subset of free licenses (like attribution only).
About Common Corpus
Common Corpus is made of six carefully curated collections:
- OpenCulture: our largest collection at 967,018,390,906 tokens, featuring public domain books, newspapers from cultural heritage repositories and open projets like Wikisource ad Gutenberg. We're developing innovative tools of OCR correction based on Pleias Models to correct historical digitization errors, while implementing advanced toxicity filtering to ensure content meets modern ethical standards.
- OpenGovernment: 579,150,518,908 tokens of financial and legal documents, including Finance Commons (from sources like SEC and WTO) and Legal Commons (including Europarl, Caselaw Access Project, Chinese Case Law), providing enterprise-grade training data from regulatory bodies and administrative sources.
- OpenSource: 283,227,402,898 tokens of high-quality code in open source from GitHub, filtered using ArmoRM to ensure only the top 80% of submissions by quality rating are included.
- OpenScience: 281,193,563,789 tokens of academic content from Open Alex and other open science reposiories, processed using vision-language models to preserve crucial document structure and formatting.
- OpenWeb: 88,517,032,065 tokens from Wikipedia (official releases from the Wikimedia Foundation on Huggingface), YouTube Commons and Stack-Exchange.
- Open Semantic: 67,958,671,827 tokens from Wikidata (official releases from the Wikimedia Foundation on Huggingface). The data has been reprocessed thanks to support and help of Wikidata and Wikimedia Germany. It includes the transcriptions of all the semantic triplets into natural language statements in over 300 languages.
| Collection | Domain | Sources |
|---|---|---|
| OpenGovernment | legal and administrative | Finance Commons (e.g. SEC, WTO) and Legal Commons (e.g. Europarl, Caselaw Access Project, Chinese CaseLaw) |
| OpenCulture | cultural heritage | public domain books and newspapers, Wikisource |
| OpenScience | academic | OpenAlex |
| OpenWeb | web text | YouTube Commons, MOSEL, Stack Exchange, CCCC |
| OpenSource | code | GitHub |
| OpenSemantic | Semantic data | Wikidata |
The first version of Common Corpus was released in November of 2024. The second version added Wikidata and detailed document-level information, including licensing and other core metadata whenever available. The third ongoing version dramatically expand the language coverage of Common Corpus beyond the US and Europe with the integration of large collection of documents in Chinese, Japanese, Arabic, Korean and Hindi.
The dataset release is accompanied by a comprehensive technical report (ICRL 2026 - oral) detailing our methodologies and data sources will accompany the release, ensuring full transparency and reproducibility.
Dataset Structure
Data Fields
identifier: unique text identifier. In many cases, this is also the link to the original resources.collection: name of one of the XX sub-collections curated for Common corpus.open type: one of the six leading collection groupings:license: sharing rights for the content either uncopyrighted (public domain, US federal public domain, CC0 on Wikidata) or various free licenses (Creative Commons, MIT, French Licence ouverte, etc.)date: date of creation of the resource where known. Due to the significance of public domain and other cultural heritage content, more than half of Common Corpus predates the 21st century.title: title of the resource when known or alternatively the filename.creator: institution publishing/collecting/curating the resource.language: automatically identified language.word_count: number of space delimited words.token_count: number of tokens as calculated by Pleias official tokenizer and Gemma-3 tokenizer for Chinese, Japanese, Arabic, Korean and few additional non-Western languages.text: full text, without formatting.
Provenance
The provenance of the datasets that make up Refined Common Corpus is detailed in the technical report [link]. Additionally, the original source URL is available in the metadata for each document for most of the dataset.
How to Use
Considerations for Using the Data
All data in Common Corpus are permissibly licensed and may be used for both commercial and non-commercial purposes.
The dataset is multilingual. The language text is included in the metadata, so data can be filtered by language. Additionally, some of the text data are historical. The year each text is written is included in the metadata, therefore it is possible to construct a dataset with a custom date cutoff if desired.
Discussion of Bias
Some of the dataset sources contain biased and toxic content, such as stereotypes about certain minoritized groups. We have removed texts which had high toxicity scores according to our toxicity classifier, Celadon, or which contain offensive terms and slurs. See our preprint for more details.
Personal and Sensitive Information
We have attempted to remove personally identifiable information (PII). We primarily use Microsoft Presidio, but make additional modifications to account for language- and country-specific considerations, such as European phone number formats.
Some small parts of the French administrative common crawl have been entirely dropped using our unreleased small reasoning model for GDPR-filtering, due to the heightened risk of transmitting identifiable indirect personal information.
Using Common Corpus
from datasets import load_dataset data = load_dataset('PleIAs/common_corpus')
Acknowledgements
The Corpus was built up with the support and concerted efforts of the AI Alliance, the French Ministry of Culture as part of the prefiguration of the service offering of the Alliance for Language technologies EDIC (ALT-EDIC).
This dataset was also made in partnership with Wikimedia Enterprise and Wikidata/Wikimedia Germany. We're also thankful to our partner Libraries Without Borders for continuous assistance on extending low resource language support.
The corpus was stored and processed with the generous support of the AI Alliance, Jean Zay (Eviden, Idris), Tracto AI, Mozilla. Generation of OCR correction at scale were performed using HPC resources from two GENCI–IDRIS grants: 2023-AD011014736 and GC011015451.
Some parts of the corpus have been built on top of other similar open science LLM community initiatives such as German-Commons, MOSEL, kl3m, AI4Bharat, Creative Commons Common Crawl. We included a new curator field to properly acknowledge this data work.
.svg.png){kind=logo}
- Downloads last month
- 60,016
Size of downloaded dataset files:
430 MB
Size of the auto-converted Parquet files:
430 MB
Number of rows:
69,907