Commit
·
eeffb50
1
Parent(s):
4262db7
update readme
Browse files
README.md
CHANGED
|
@@ -1,19 +1,37 @@
|
|
| 1 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
license: agpl-3.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
-
|
| 6 |
|
| 7 |
-
|
| 8 |
-
German (de), French (fr), Italian (it), English (en), and Luxembourgish (lb).
|
| 9 |
|
| 10 |
-
|
| 11 |
|
| 12 |
-
|
| 13 |
-
pip install transformers floret
|
| 14 |
-
```
|
| 15 |
|
| 16 |
-
##
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
```python
|
| 19 |
from transformers import pipeline
|
|
@@ -34,3 +52,42 @@ face à une opportunité."""
|
|
| 34 |
langs = lang_pipeline(text)
|
| 35 |
print(langs)
|
| 36 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
+
library_name: transformers
|
| 3 |
+
language:
|
| 4 |
+
- fr
|
| 5 |
+
- de
|
| 6 |
+
- en
|
| 7 |
+
- it
|
| 8 |
+
- lb
|
| 9 |
license: agpl-3.0
|
| 10 |
+
tags:
|
| 11 |
+
- language-identification
|
| 12 |
+
- multilingual
|
| 13 |
+
- historical
|
| 14 |
+
- impresso
|
| 15 |
---
|
| 16 |
|
| 17 |
+
# Model Card for impresso-project/language-identifier
|
| 18 |
|
| 19 |
+
## Overview
|
|
|
|
| 20 |
|
| 21 |
+
`impresso-project/language-identifier` is a multilingual language identification model fine-tuned for use on historical newspaper content. It supports **German (de), French (fr), Italian (it), English (en), and Luxembourgish (lb)** — the core languages of the [Impresso Project](https://impresso-project.ch), which focuses on analyzing historical media across national and linguistic borders.
|
| 22 |
|
| 23 |
+
This model has been adapted for short, OCR-noisy and fragmentary inputs typical of historical digitized texts.
|
|
|
|
|
|
|
| 24 |
|
| 25 |
+
## Model Details
|
| 26 |
+
|
| 27 |
+
- **Model type:** Language identification
|
| 28 |
+
- **Interface:** Hugging Face `transformers` pipeline
|
| 29 |
+
- **Languages supported:** fr, de, en, it, lb
|
| 30 |
+
- **License:** AGPL-3.0
|
| 31 |
+
- **Developed by:** UZH, Switzerland
|
| 32 |
+
- **Training data:** Historical newspapers from the impresso corpus and related sources
|
| 33 |
+
|
| 34 |
+
## How to Use
|
| 35 |
|
| 36 |
```python
|
| 37 |
from transformers import pipeline
|
|
|
|
| 52 |
langs = lang_pipeline(text)
|
| 53 |
print(langs)
|
| 54 |
```
|
| 55 |
+
|
| 56 |
+
## Output Format
|
| 57 |
+
|
| 58 |
+
The output is a single dictionary with the predicted language and confidence score:
|
| 59 |
+
|
| 60 |
+
```python
|
| 61 |
+
{
|
| 62 |
+
"language": "fr",
|
| 63 |
+
"score": 1.0
|
| 64 |
+
}
|
| 65 |
+
```
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
## Use Cases
|
| 69 |
+
|
| 70 |
+
- Preprocessing for OCR and NLP tasks on historical corpora
|
| 71 |
+
- Document and segment-level language tagging
|
| 72 |
+
- Filtering and sorting multilingual newspaper archives
|
| 73 |
+
|
| 74 |
+
## Limitations
|
| 75 |
+
|
| 76 |
+
- Works best on **sentence- or paragraph-length** texts
|
| 77 |
+
- May struggle with code-switching or OCR-degraded text that mixes languages
|
| 78 |
+
- Primarily optimized for **Impresso-like sources** (19th–20th century newspapers)
|
| 79 |
+
|
| 80 |
+
## Installation
|
| 81 |
+
|
| 82 |
+
```bash
|
| 83 |
+
pip install transformers floret
|
| 84 |
+
```
|
| 85 |
+
|
| 86 |
+
## Contact
|
| 87 |
+
|
| 88 |
+
- Website: [https://impresso-project.ch](https://impresso-project.ch)
|
| 89 |
+
|
| 90 |
+
<p align="center">
|
| 91 |
+
<img src="https://github.com/impresso/impresso.github.io/blob/master/assets/images/3x1--Yellow-Impresso-Black-on-White--transparent.png?raw=true" width="300" alt="Impresso Logo"/>
|
| 92 |
+
</p>
|
| 93 |
+
|