diff --git "a/data/test_audiocaps.raw.json" "b/data/test_audiocaps.raw.json"
new file mode 100644--- /dev/null
+++ "b/data/test_audiocaps.raw.json"
@@ -0,0 +1,761 @@
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YH4qi8Ul6v0_30.wav", "text": "A man talking as an infant is crying followed by a man humming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GE1aZSnPr2Q_0.wav", "text": "A man laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jj2RyNDj7no_440.wav", "text": "Birds chirping as insects buzzing followed by a man talking in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/gbi6MxPf3hA_30.wav", "text": "Winds and ocean waves crashing while a chime instrument briefly plays a melody"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tTB0BK39JI8_40.wav", "text": "Bells ring followed by clanking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Hdxfbpnd2-8_140.wav", "text": "A man talking then whistling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yL3gKa6YLoM_40.wav", "text": "A man speaking followed by snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wSHzVxdMiTo_30.wav", "text": "A pig squeezing and male speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6dLkgq9EKPE_30.wav", "text": "A man speaks with some faint clicks and humming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/gr5Zss89yLQ_30.wav", "text": "A woman speaks, and a motor vehicle revs its engine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/t4prXmPwthg_150.wav", "text": "Vibrations from a running sewing machine with a woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xYwpABpZed4_370.wav", "text": "A woman speeches while cooks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/emGPabOePzA_30.wav", "text": "Laughing and speech in a slowed speed"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Yk274Wr5iIE_60.wav", "text": "Wind and water splashing and trickling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/XrJcmftCY04_180.wav", "text": "A crowd of people applauding and cheering"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WCYfCfW9NA0_230.wav", "text": "High pitched humming with whooshes of spinning helicopter blades"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fPqj3nnwQOI_270.wav", "text": "Ocean waves crashing and water trickling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yVVLq4ao1Ck_30.wav", "text": "Insects trill in the background, while birds chirp and flies buzz"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bygBWUkpaC8_270.wav", "text": "A man speaking with light wind followed by brief silence then birds chirping."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/qF72bT878gw_14.wav", "text": "Humming from an engine with wind blowing hard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/g_P29ucKj78_30.wav", "text": "A race car accelerating and driving as plastic rattles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6OlHuvJR_Dk_30.wav", "text": "Helicoptor flying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bX2vDaHL26U_60.wav", "text": "A car skidding its tires and racing away"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bA5zPFSFZAA_7.wav", "text": "Static and beeping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IPfaRF76gVU_13.wav", "text": "A truck engine driving by while emergency sirens wail as a crowd of people talk followed by a man speaking over an intercom"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HVz-FJBf_iM_10.wav", "text": "A toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PYP-r0nvbFk_30.wav", "text": "Wind blows and a man speaks followed by a goat bleating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/uY4fe5DT1gI_30.wav", "text": "A person is typing on a computer keyboard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-EaZ7EJJUl0_50.wav", "text": "A person speaks with distant humming and nearby clinking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KvrcRMfFzOE_150.wav", "text": "A helicopter is in flight with loud blades moving fast"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/c6YJgZ3qzOw_380.wav", "text": "A toy helicopter flying followed by wind blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8nUqSYC66mI_30.wav", "text": "Several people cheer and scream and speak as water flows hard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9zstu_IfAm4_30.wav", "text": "A truck horn honking as a large truck engine drives by while a crowd of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Y3lNEe-ZGF0_30.wav", "text": "A clock ticking followed by a wooden latch opening"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Vkbp8VmL3pM_40.wav", "text": "An infant crying as clothes rustle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BLMWD6fxhpo_22.wav", "text": "A grunt from a person followed by footsteps and oinks from a pig with a single squeal"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CBwXKOpJY_o_30.wav", "text": "A woman and young boy talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GIOApFAWDOc_160.wav", "text": "Birds chirping and tweeting followed by light wood tapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7JWHbs3gu1w_180.wav", "text": "A train roars and honks it horn"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Bn4lc01q9vE_30.wav", "text": "Water splashing followed by a woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/eRU-rABp8nk_30.wav", "text": "Two men talking as a motorcycle engine runs idle and a group of people laugh"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/f8WPf5F22xI_40.wav", "text": "Sneezing followed by a ticking clock and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Qt0_xTadAT0_27.wav", "text": "A group of frogs croaking as plastic flutters in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lTfNLKEy1RU_120.wav", "text": "Repetitive faint snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9PN4gyxpH2M_40.wav", "text": "A man and a woman talking then the crinkling of paper"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MkbP_8zJwXU_40.wav", "text": "A bird chirping as water lightly trickles and wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/o_3MDLl_aH0_20.wav", "text": "Loud explosions and burst that echo"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/R_g4RpU9mO0_30.wav", "text": "A motorboat engine running idle then accelerating as water trickles followed by water spraying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ARFFw0e_jig_20.wav", "text": "A loud distorted belch followed by a series of burping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/paetCbEqp2w_60.wav", "text": "Some light tapping on a computer keyboard and a baby crying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kLYCjD6vWI4_30.wav", "text": "Hissing of a steam engine train"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4Ujigme2IxY_30.wav", "text": "An engine humming and brakes squealing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zIgGMlZENTs_30.wav", "text": "A duck quacks followed by a man talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Lbken4JCr94_10.wav", "text": "Rolling thunder with lightning strikes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/14ekd4nkpwc_28.wav", "text": "An infant crying followed by a man laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CYUlLTKoa1Y_80.wav", "text": "A woman speaks as a machine whirs and creaks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hiJB_95IWiE_190.wav", "text": "Male speech and then scraping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/--0w1YA1Hm4_30.wav", "text": "High pitched speaking and laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/q4YFJA5pFXc_110.wav", "text": "A helicopter engine whines as it starts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6NBPiArs2-w_80.wav", "text": "Footsteps followed by rapid gunshots and people speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Kel-hfZ_9h8_10.wav", "text": "Crunching and scraping followed by a child and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/41D0yXSBqfI_30.wav", "text": "Pigeons cooing and bird wings flapping while gravel shuffles and wood clacks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_z6pymOet7g_50.wav", "text": "A man talking followed by a toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CMNlIW6Lkwc_12.wav", "text": "A man is talking while plastic clacks followed by a popping explosion then a group of people talking and laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pTJKJxaheI8_30.wav", "text": "A man speaks with faint ticks of a clock"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BUAPM4D3-h8_290.wav", "text": "Women, child talking and birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2EsxcKe1A4w_28.wav", "text": "A large dog barks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6ZFU4PqXmoI_6.wav", "text": "A construction vehicle engine running and water splashing as wood crackles and snaps followed by electronic beeping and a man talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MdlEswBfZMQ_70.wav", "text": "One toddler crying with several others laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AKHZMg9ba30_30.wav", "text": "Clip-clop of horse while man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KJKHDKKW3XU_5.wav", "text": "Gurgling and splashing water"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/G0IsabU5hn4_3.wav", "text": "Wind blowing heavily into a microphone as ocean waves crash and leaves rustle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1e98HeU9Vrg_20.wav", "text": "Waves crash on a beach while the wind blows"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AUmY0YRAFQE_0.wav", "text": "Sirens ring as a vehicle speeds past"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/J0yeFeKvIt8_30.wav", "text": "Wind blowing and leaves rustling followed by plastic clicking and wind blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/vigslb0kClE_100.wav", "text": "Goats bleating and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2JutOgAnqWA_120.wav", "text": "A tool motor is running, a rattling sound occurs, then humming also, then soft squealing occurs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Bz9Y5nZK3eo_21.wav", "text": "Typing on a keyboard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ivg_q4t-3w0_1.wav", "text": "White noise followed by brief, muffled speech and rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/R4fXcbWFhJg_10.wav", "text": "Horses run in the distance then a man speaks and a woman cheers then talks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lh801oHGtD4_30.wav", "text": "A man talking as an electronic device hums followed by a door closing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/VeCSHwtkBZU_30.wav", "text": "A police siren going off"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bAqgL5dCQOE_30.wav", "text": "Rain falling with distant hums of traffic"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hV4bDCBDCy0_90.wav", "text": "Humming of a large truck engine with blowing wind"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jinJkonlrWc_330.wav", "text": "An engine running and muffled speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BwnGxJD9xh8_30.wav", "text": "Speech and a goat bleating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/j_NSuPnx5LA_40.wav", "text": "Dialing phone followed by clack sound"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3fomsZXG3aM_3.wav", "text": "A vehicle engine running idle then revving after two metal clicks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kXjzsroVTtw_20.wav", "text": "Birds chirp and sing in the background, while an adult male speaks and crunching footfalls occur"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rINmxSXMR-s_30.wav", "text": "A duck quacks and splashes and a human chuckles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pWQeV08kYR0_10.wav", "text": "An ambulance sounds the siren"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Qv1HXaT-28U_4.wav", "text": "Liquid spraying and splattering on a hard surface as kids laugh in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0qbHT34qTZE_50.wav", "text": "Male yelling and multiple gunshots"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/VZLZ08k3YeA_60.wav", "text": "Someone snores and a man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ljrL7Cb-jr8_30.wav", "text": "Humming and vibrating of engines"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fYTZVxQ8LJk_0.wav", "text": "A woman is speaking, and a child speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Dn3buZWMzwY_20.wav", "text": "A man and a kid laughing followed by another man speaking as a person snores"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4xrL4TSgHwU_30.wav", "text": "A vehicle engine stuttering while starting up then running idle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fGGYeXR_LS8_30.wav", "text": "Whistling is ongoing while an adult male speaks, and water is pattering in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/K8-b0VtNOqA_4.wav", "text": "Horses snorting then neighing as footsteps shuffle and people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/k4XyfaWVLEY_30.wav", "text": "A vehicle engine idles and birds chirp in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EvZ3jOMYWxk_30.wav", "text": "A woman is speaking in a quiet environment"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sl_Pxpc7beo_30.wav", "text": "A vehicle is honking the horn"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jf4iyQPJSvk_20.wav", "text": "Water flowing and trickling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/paf0nyjg1Js_110.wav", "text": "A woman talking followed by a plate rattling as food and oil sizzle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/huMLK0oA3L8_110.wav", "text": "A man talking then whistling as guitar music plays"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/onBZOH88OYs_30.wav", "text": "Repeated bursts of spray"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HxZADVzNIqs_30.wav", "text": "Birds chirping and water dripping with some banging in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FKaJsvcyHTk_0.wav", "text": "A baby grunts and begins crying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GPj8h-WcjWs_21.wav", "text": "Humming of an engine with a woman speaking over a loudspeaker"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kHIe4CfaccQ_40.wav", "text": "Continuous basing and rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_BSmz3SEW1w_30.wav", "text": "Pigeons coo and rustle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xbLW9Wt1Jsg_70.wav", "text": "Some rustling with short bursts of vibrations from a sewing machine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/H7-orYrKBeo_70.wav", "text": "A group of people talking as an infant cries while a muffled bus engine runs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YQSuFyFm3Lc_230.wav", "text": "Clattering of a train is ongoing, a railroad crossing bell rings, and a train horn blows"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-SkjbQVgJ0M_30.wav", "text": "A car passes by shortly after a man starts talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CeRoaEcqUgM_30.wav", "text": "A motorboat driving by as water splashes followed by wind blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ztSjcZNUY7A_30.wav", "text": "A woman talks as a baby cries"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1Og2TJ3bXW0_20.wav", "text": "An aircraft engine humming followed by plastic clanking then an aircraft engine slowing down"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/o7-X8DAToGc_170.wav", "text": "A vehicle driving by with tires briefly skidding and accelerating then slowing down"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sI7_ycEYzAY_30.wav", "text": "A clock ticks repeatedly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/k1QxQ4jJaEQ_340.wav", "text": "A man talking as an idle truck engine runs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/B8rdur4aams_30.wav", "text": "Rustling and shaking followed by a horn and gusts of wind"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/e9MWXS34o48_40.wav", "text": "A person breathes heavily and sneezes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bUTOsLXYyxg_220.wav", "text": "Man speaking followed by group laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ba92IrXFvJo_30.wav", "text": "Cracking and humming from an engine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Zsf2YvJfCKw_20.wav", "text": "Water trickling followed by a toilet flushing then liquid draining through a pipe"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Z_smJ66Tb3c_30.wav", "text": "A man talking as pigeons are cooing and birds are chirping with bird wings flapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1Uh74_rZ72k_30.wav", "text": "Wind blowing with passing distant traffic and some rumbling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Rk-ujWKzPuc_280.wav", "text": "Hard rain is falling and splashing on a surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OMGHnJV0l2U_350.wav", "text": "Scraping and banging"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/C4JwGJQ2dUA_30.wav", "text": "A man yelling very loudly while speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0jGH7A_hpBM_30.wav", "text": "An adult male speaks, followed by another adult male speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5K1mISHwggI_30.wav", "text": "A man speaks with others speaking in the distance and traffic passing with a few horns honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MTaLknnq4wc_60.wav", "text": "Whistling with some distant humming and a girl singing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dmUOSyPXkUw_200.wav", "text": "Compressed air and steam releasing with a man faintly talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sVYTOURVsQ0_0.wav", "text": "Water trickling with man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_duNX6Vyd6g_30.wav", "text": "A motorboat drives on water quickly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PVvi2SDOjVc_0.wav", "text": "A vehicle is running followed by the horn honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GOD8Bt5LfDE_100.wav", "text": "A toy train running as a young boy talks followed by plastic clanking then a child laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ii3Geza3hAU_210.wav", "text": "Vibrations from a sewing machine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jOYvIISk--4_110.wav", "text": "A man speaking followed by a faucet turning on and off while pouring water twice proceeded by water draining down a pipe"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TwR8BA6buMI_30.wav", "text": "A piano playing as a clock ticks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EBCH7TPgiPc_0.wav", "text": "Engine revs very high for a while, then begins to slow down"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/S8k47ME-YT4_100.wav", "text": "Plastic cranking as rain heavily pours while a stream of water rapidly flows"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/o3mZR8OvPko_30.wav", "text": "A racing vehicle accelerating and revving while tires skid and squeal"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xIztYnMIWUA_30.wav", "text": "An old fashioned telephone ring occurs repeatedly, and buzzing is present in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IJ6pm5Kns8A_170.wav", "text": "Woman speaks, girl speaks, clapping, croaking noise interrupts, followed by laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NDaVSIJaXVs_410.wav", "text": "Aircraft engine and loud background roar"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/35b9BSmN5JM_1.wav", "text": "A wooden thud as an idle car engine runs then accelerates"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2sZhC_mKeic_30.wav", "text": "A cat meowing followed by metal clacking as a water sprinkler sprays in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LvhvAA11oxE_10.wav", "text": "A vehicle engine repeatedly revving as tires skid while a man talks over an intercom and wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SL3wB5sDcdw_350.wav", "text": "Leaves rustling as a leaf blower operates and wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MTIF_l_8d4Q_0.wav", "text": "A baby is crying and a child speaks in the background, then an adult female speaks and the baby sobs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TSdAJWJ-tW0_12.wav", "text": "Two vehicles drive by as a man is speaking in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/gbtcDoh0q3c_130.wav", "text": "Metal clacking followed by a man talking then a metal bang as footsteps shuffle on dirt and a group of men laugh"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_9mgOkzm-xg_120.wav", "text": "A man speaks as food sizzles followed by some cracks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sbW7XwwUtSU_23.wav", "text": "A clock ticking repeatedly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/d1tL-9BILy8_30.wav", "text": "Pigeons cooing followed by camera muffling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Pg2cWEnEEvc_60.wav", "text": "Burping noises followed by tongue clapping sounds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4abZbau8tZo_14.wav", "text": "Waves roll in and a man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Lxu-3_h4kc4_0.wav", "text": "Burping and then laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UE3XnVFodMI_230.wav", "text": "Loud applause and woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nq0BF9zGkzg_30.wav", "text": "An animal cries, followed by a crash, then two people speak with one another"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mYQrjcYNrW0_20.wav", "text": "A bus horn honking as wind is blowing into a microphone before a bus drives by"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/w_Utn3CwAXE_50.wav", "text": "A toilet flushes followed by a metal latch clanking then a door slowly creaking open"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AFgGoY8Ihhg_30.wav", "text": "Wind blowing and a siren rings"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2RjqBRzmxaM_30.wav", "text": "Music with rain falling and a woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yrxa6_P2I80_30.wav", "text": "Several birds chirping and pigeons cooing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Dc2WEiRk0rA_30.wav", "text": "Hissing spray is occurring, and liquid is pattering on a surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mUGmCSNETcg_310.wav", "text": "Women talk while food fries"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hJtOGmN_KVw_140.wav", "text": "Plastic crinkling and paper sliding as a man talks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/67BsqRkh-dU_10.wav", "text": "A toilet is flushed"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NwoBDrTlbTI_30.wav", "text": "A gerbil squeaking during audio static"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/D9tinq3RMpU_30.wav", "text": "A large motor is running smoothly, water is splashing, people are talking in the background, and an adult male speaks in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0G7rb74R-2A_30.wav", "text": "A man speaking on a microphone as a crowd of people laugh followed by dinner plates clacking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/30D1tqNFHMc_30.wav", "text": "A loud white noise and then some beeping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Me4npKmtchA_270.wav", "text": "Bubbles gurgling and water spraying as a man speaks softly while crowd of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/52IxrdTxGs4_30.wav", "text": "An explosion and heart beating as instrumental music plays followed by a man talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/c3nlaAkv9bA_230.wav", "text": "A male voice talking and the bleat of a goat"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9vZDsGjyh5M_40.wav", "text": "Vehicle motor running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/s_EWjoiVfzo_310.wav", "text": "A sewing machine operating several times followed by metal clanking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/h3UhoHIMfpw_30.wav", "text": "Some rustling with a very high toned beep"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hmd6pa2e_rs_30.wav", "text": "A bus engine is running, and people are talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ErxgH5g3Kx0_30.wav", "text": "Hissing and buzzing of insects with a horse neigh and clip clops"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2KEfkDO6hlA_30.wav", "text": "Vehicle accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2ErfX6ZT5pM_0.wav", "text": "A toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4pv3w--cRrA_16.wav", "text": "An electric motor runs then a person speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jjHIINDfE1c_20.wav", "text": "A tractor driving by as a car horn honks while wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/V4PLSw_WzVw_10.wav", "text": "A vehicle engine revving as tires skid and squeal"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_z-bidQYVao_22.wav", "text": "Male rapping and screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/A61Mry8zBwE_30.wav", "text": "Applause from a crowd with distant clicking and a man speaking over a loudspeaker"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LB6CZ0x-kns_0.wav", "text": "A loud metal clank followed by pressurized air releasing as a truck engine runs idle before accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2KR0C5ysO8o_30.wav", "text": "An engine revving continuously"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KVAIaRPry24_50.wav", "text": "Flying insects are buzzing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IsUG5SKWNZA_30.wav", "text": "A woman whispering then talking as an infant cries followed by a man yelling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/27HIamF8pKo_110.wav", "text": "Railroad tracks clickety-clack as a train horn blasts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3hzy-FL24no_17.wav", "text": "A stream of water flowing and trickling as a small motor hum in the background followed by camera muffling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/RrmBGjJqlEo_400.wav", "text": "Clicking on a keyboard with a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Tgxst7Ft9js_20.wav", "text": "Sanding and filing then a man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-aYumc8KoXg_60.wav", "text": "A vehicle is driving, and a horn is honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8VOibo9Q_Dc_30.wav", "text": "Water splashing and birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Cbe2B6ohBpw_30.wav", "text": "Continuous quacking and rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OFVzrakJhbw_30.wav", "text": "A woman laughing followed by the bleating of sheep and a soft breeze"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/D96OO7nYYsg_30.wav", "text": "A vehicle accelerating then revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KVbmN9ZRg5Q_220.wav", "text": "Loud hissing followed by a loud continuous horn and brief clanking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/gkWd1HugK2w_30.wav", "text": "Pigeons coo and flap their wings"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kx6Rj4MDIAw_30.wav", "text": "A cuckoo bird coos followed by a train running on railroad tracks as a bell dings in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NJEPbGVBJIQ_10.wav", "text": "A man speaks followed by a vehicle revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8IdCiapDYCU_20.wav", "text": "A few electronic-sounding animal noises followed by barking from a small dog"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rE6BJ0Bo4w4_200.wav", "text": "A woman speaks, and faucet water runs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1vCYiVvZ7VE_30.wav", "text": "A woman speeches"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/h0M4RS8p_mo_0.wav", "text": "A man laughs followed by a baby crying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WUyeFOyKIg0_250.wav", "text": "Wind and people speaking and a car passing by"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0Rpjl1AO-P0_30.wav", "text": "A car engine idling then starts to rev shortly after"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/n74IYuCe_ms_110.wav", "text": "A stream of water rushing and trickling followed by a young man whooshing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Mj_BO-iK1G4_310.wav", "text": "Vibrations from a sewing machine with clicking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Z0IrCa4MvOA_30.wav", "text": "Water flowing through pipes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zoctgurhvHE_70.wav", "text": "A man speaks with some clicking as wind blows and a distant motor hums"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WqXFAY4k79s_230.wav", "text": "Several loud pops followed by one woman laughing and another speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lX3k5p2I_g0_140.wav", "text": "Man talking in the wind and someone yells in the background while an engine makes squealing and air puffing sounds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zEaGx6an4es_270.wav", "text": "A sustained and coarse drilling noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FhimNYClv40_40.wav", "text": "Sirens and a humming engine approach and pass"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4KObP7cREWw_40.wav", "text": "Engine running and revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/js4dr5JusdM_30.wav", "text": "A woman and man talk on the phone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Flk-X0gwjF4_510.wav", "text": "Birds are chirping and singing, rustling occurs and an adult male speaks briefly in the foreground, and an adult male speaks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/79XDcI6xZm0_30.wav", "text": "A crowd chants, an adult male speaks in the foreground, a member of the crowd shouts, and the adult male continues to speak"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jXkLS_QzUrI_29.wav", "text": "A small motor is running, five hisses and a meow occur, then five more hisses and a meow occur"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2JV3emH50XU_0.wav", "text": "A vehicle accelerating then driving by as gusts of wind blow and leaves rustle in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nLtNjMimLE0_480.wav", "text": "Running and splashing water"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BQ-r9mEHssU_17.wav", "text": "A woman laughs and talk as a hose clip-clops"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/B4SZwi9Ce3o_30.wav", "text": "A man talking and plastic clicking as a vehicle engine accelerates"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Smdj6JFB9MQ_350.wav", "text": "Female speech, a toilet flushing and then more speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/qakN0JNbpcU_170.wav", "text": "A man is speaking and a fly is buzzing around"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/G3YO2unWz7k_290.wav", "text": "Sputtering of an idling and revving engine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/l5KdHAWwJCw_30.wav", "text": "Ticking of a clock"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SoO1HhaEc9Q_30.wav", "text": "Rhythmic squeaking is ongoing, pigs oink, an adult male speaks, and music plays faintly in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EUZaxaWqhwg_40.wav", "text": "A slap occurs and an aircraft engine runs as a man and child talk"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5ye0X5saadg_30.wav", "text": "Helicopter running, speech on a radio, and then gunfire"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/VQnmlf2OsUg_50.wav", "text": "A toy helicopter flying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4sb9jN0SgTM_26.wav", "text": "A motor vehicle engine runs and revs, then it stops and an adult male speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fwhkCnOeyC0_350.wav", "text": "A man and a woman talking followed by a bell ringing and a cat meowing as a crowd of people applaud"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CxaPpRJRkn0_0.wav", "text": "Male snoring loudly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/C_ga5m6nOhI_30.wav", "text": "A train running on train tracks while a steam engine horn whistles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/P12nvSpKXcs_24.wav", "text": "Birds are calling and many bees are buzzing, then a child speaks briefly and rustling footfalls and rattling occur"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HkbCUN4V3TU_20.wav", "text": "A girl screams and people speak"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/I4HpYGMMsz4_170.wav", "text": "A man speaking with slapping on a wooden surface as compressed air hisses in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9z8XIRyUq9Q_30.wav", "text": "A woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4IeDBwyQ9ZQ_190.wav", "text": "Woman speaking, plastic container opening"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/uhSDBwVrEdo_40.wav", "text": "The wind is blowing, and a person is whistling a tune"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9BukzlPNqC8_60.wav", "text": "Soft drilling noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tIM-H2rdq8U_60.wav", "text": "A gun firing several times followed by a revolver chamber spinning and metal clanking as a man talks and a person grunts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AJtNitYMa1I_2.wav", "text": "Food and oil sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/c0V_HAul7rI_30.wav", "text": "People laugh and speak"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dYvL6uEMl6E_60.wav", "text": "A helicopter flying followed by wind heavily blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mWqH2xwjkYA_50.wav", "text": "A baby laughs and a woman speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pHYkWkZ4guE_320.wav", "text": "Scraping and hissing with some light knocks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PWjEfOkb6ro_21.wav", "text": "Water running with distant speech in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/RdC8cviN6Bs_30.wav", "text": "Rain falling followed by fabric rustling and footsteps shuffling then a vehicle door opening and closing as plastic crinkles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/m_U506sf9p4_170.wav", "text": "A woman talking followed by plastic thumping then metal clanking as food and oil sizzles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NeZerAPXR-A_0.wav", "text": "Woman laughing and talking with another man and woman while something is clanging"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0AsXkZkqelg_30.wav", "text": "A loud bang followed by an engine idling loudly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0a9wVat2PWk_230.wav", "text": "A train sounds horn and travels"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lfAFQ0-wDJU_30.wav", "text": "Wind and surf with water splashing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-DmjkgWa-rw_10.wav", "text": "A bell ringing repeatedly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/aMhu5eMQAsI_140.wav", "text": "Humming from a large engine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3wrdPAeqjVI_80.wav", "text": "A man talking followed by a camera muffling and footsteps shuffling then wood lightly clanking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dZDgJzGtLLU_290.wav", "text": "Water splashing with some rustling and wind blowing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fWvWhLJ5Fow_30.wav", "text": "Running then spraying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oiIi6H83Y38_0.wav", "text": "A motorbike engine starting up then revving as a man talks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PRUfwpmYwJ8_30.wav", "text": "A man talking as compressed air is spraying with wood clanking in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wNiYSYJXssA_460.wav", "text": "People are speaking followed by crowd noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UhCzD6EBJBU_180.wav", "text": "A power tool drilling followed by metal clacking then a man talking proceeded by a door closing shut"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CfxWJ1Qoufg_10.wav", "text": "A man talking as paper crinkles followed by plastic creaking then a toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7XUt6sQS7nM_30.wav", "text": "Sheep baaing with young female voice and wind gusts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/83j4GgHXTLE_210.wav", "text": "A kid laughing and screaming with a man chuckling and woman whispering"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AR8-MVl_Mf8_2.wav", "text": "A man screaming followed by slamming against a metal object proceeded by a man talking cardboard tumbles on a hard surface as a truck engine runs idle in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HeEa1GZpUGI_30.wav", "text": "Several gunshots followed by a click and glass breaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Xf5LjaE_JQ0_30.wav", "text": "Traffic sounds going by as man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UAmDLPjNyMg_40.wav", "text": "An aircraft flying as a person briefly sniffs aloud while wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GGgQR7aIofY_40.wav", "text": "A horn beeps a tune, and a large motor vehicle engine is running and gets louder"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hGWarNR6xmg_110.wav", "text": "Repetitive loud hissing with some static"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yLu4b01t53k_240.wav", "text": "The loud idling of a machine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rjUrB1WUpcI_14.wav", "text": "Water drains with gurgling noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JBWJQCS4SvA_30.wav", "text": "Wind noise on microphone with waves splashing in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mSF_FqBtRPs_20.wav", "text": "A telephone ringing with loud echo"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LVvS3s9dFKw_30.wav", "text": "A man speaking while water runs in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ls1zyPjs3k8_280.wav", "text": "A bleeping noise followed by a static noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/v7BaYF0kagM_310.wav", "text": "An engine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NmmbNqmsPaY_10.wav", "text": "An engine idling with bells ringing in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KnXNy5Q6YS4_230.wav", "text": "Man speaking while insects buzz around"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GuizRlAQ8qQ_50.wav", "text": "High powered vibrations from a power tool"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1DKLyH3FixM_100.wav", "text": "Nature sounds with birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1PvMtRIlZNI_0.wav", "text": "Gurgling and splashing water"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SE_3nszEw7o_22.wav", "text": "Hissing and clinking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/L_CNz9Vrtkw_100.wav", "text": "A man speaks followed by a crowd applauding"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ItS07xtdi4s_30.wav", "text": "Fire igniting as a motor runs followed by an electronic beep and vehicle engines running idle then car horns honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7MLERaOgK_Y_200.wav", "text": "A girl sings while a sewing machine stitches and then speaks when it stops"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/p_BB_rJaF7Q_0.wav", "text": "Male speaking softly with nature sounds in background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YNDKuNINDOY_30.wav", "text": "An emergency siren ringing with car horn honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ZsTZ7jqbd9M_170.wav", "text": "A man talking during footsteps on leaves as birds chirp in the background."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7P0N61TVOxE_150.wav", "text": "A band plays horn and string instruments while waves crash nearby"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KnsKf9KoNds_210.wav", "text": "A man speaks nearby and another speaks over the phone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YIqpIjjee00_30.wav", "text": "A flush occurs with water splashing and gurgling, and then water runs steadily"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FXdoNvmrYxo_280.wav", "text": "A young girl talking followed by a man talking then a young boy softly talking as birds are chirping in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ch0LMmhBUg4_260.wav", "text": "A man speaks with some distant hum and a child speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-mb4Fw4Z0xg_17.wav", "text": "A car engine accelerating and then laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/y1a8PntuXYw_0.wav", "text": "Mechanical humming followed by a flushing toilet"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kF1KWybdRpM_60.wav", "text": "Aircraft engine roars as it approaches and then recedes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/cPiSd5nJLrI_30.wav", "text": "Screaming and a big explosion and then screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BDpU2Qh77NE_30.wav", "text": "A bird making vocalizations from a distance followed by several people whispering nearby"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TOaQMYc79Mw_40.wav", "text": "A vehicle struggling to start with some clicks and whines"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SNy_axSCoyw_30.wav", "text": "Mechanical sound with clock tick-tocking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IvfaKPDWC00_160.wav", "text": "A siren ringing with a vehicle speeding closer"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/C8kR19NvynA_30.wav", "text": "A man speaking continuously"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BL8ksJ0sTXk_260.wav", "text": "An engine is idling followed by a male speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/350OCezayrk_30.wav", "text": "An engine being turned over and then started"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WOywdRmySs0_190.wav", "text": "A man talking followed by plastic crinkling and crumpling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Rp4Ct_TQvAM_520.wav", "text": "Rain falling as a machine motor runs followed by a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wnqUgK_-fo4_0.wav", "text": "Several loud pops with people cheering"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JsoBpL86R5U_10.wav", "text": "A woman talking then a child talking followed by a man talking then a goat yelling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BzHTqyX69pI_60.wav", "text": "Wood sawing and scrapping followed by a woman speaking while pigeons are cooing and bird wings are flapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JC2ZrXzCX4Y_110.wav", "text": "A person snores while people communicate"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-FW109cbv0g_30.wav", "text": "Speech and then a sneeze and laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bpv_LneHmfU_10.wav", "text": "An engine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PZBUdlKwX04_21.wav", "text": "A plastic clack followed by a man talking as a stream of water rushes and a crowd of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ggN4-K5AgoM_90.wav", "text": "A toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bIV3bJZpkgA_30.wav", "text": "A helicopter is in flight"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ilspW7JRjAg_50.wav", "text": "A motor vehicle engine is revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/S0SQyFXbqF8_30.wav", "text": "A child talking then laughing with a man as an animal gurgles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EbpOXac13yo_30.wav", "text": "Loud bus roaring and voices"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mGa2JgAiKV8_30.wav", "text": "Young man speaking with slight movement sounds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Hg6HxylRGDo_29.wav", "text": "Emergency sirens wailing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/maVYiednkSg_60.wav", "text": "A man speaks with spraying and vibration in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/cFoXRmGgIME_20.wav", "text": "A train horn honking as a train is running on a railroad track and wind is blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/C9NC7wJ7C3w_30.wav", "text": "A woman speaking continuously"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ZYWCwfCkBp4_30.wav", "text": "Rustling and sanding on wood with distant music playing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3ejndVEAcmQ_11.wav", "text": "A cat meowing and hissing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UV1kdjwpy6U_17.wav", "text": "An engine vibrates followed by a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4SZ7JXDCNps_210.wav", "text": "A car engine revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/P4qd8uodw_M_0.wav", "text": "Muffled man speech with tapping and clicks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mW1EpJYcy_E_70.wav", "text": "A motorbike accelerating and revving as wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/E6FH_xp3I54_60.wav", "text": "A man is speaking followed by birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ujje3lSabsg_0.wav", "text": "Soft, rapid snoring is ongoing, and faint speech occurs in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YEYeQ0lIkBQ_0.wav", "text": "A man talking then laughing as a ducks quack"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/eqcdsdLz954_21.wav", "text": "A synthesized whooshing followed by an explosion and fire crackling while gusts of wind blow"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Fi4-IqJo2xQ_20.wav", "text": "Humming and rattling from a revving engine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ific_gRalg0_30.wav", "text": "A series of loud metal hammering as liquid drains down a pipe followed by a metal chain rattling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DrCm-HpX67k_180.wav", "text": "Opening and closing followed by tapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PTyFYxXdut4_370.wav", "text": "A woman speaks as water splashes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lVr-PxhZo8s_40.wav", "text": "Car engine idling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3Sml1wHcuxo_40.wav", "text": "A train horn goes off loudly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_iUX8CibElk_120.wav", "text": "Humming sounds from a distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/K-7Y8yhcUiw_60.wav", "text": "An animal snores"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HdPSebdDxe4_90.wav", "text": "A male speaking and then growling followed by more male speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hrv6fwnmBkY_28.wav", "text": "A dog whimpering followed by laughing and barking."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lq9RvAA4mqY_180.wav", "text": "A man talking then clearing his throat as food sizzles and a metal pot clangs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PO8Nu3F8mkA_30.wav", "text": "Several pops and bursts of fireworks exploding"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MBP4RcnwGZw_40.wav", "text": "A woman talking then laughing as a crowd of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/54eRRbCtPn8_30.wav", "text": "An adult female speaks in a quiet environment"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ZY4aGEniU_E_90.wav", "text": "Food and oil sizzling followed by steam hissing as a man talks and light rock music plays in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/D1Sy7kRoaR8_30.wav", "text": "A woman speaks via a public address system"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/iOCpICiu4LA_320.wav", "text": "Insects buzzing and birds chirping as a man talks followed by a loud plastic popping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dkiwn2FdDVw_0.wav", "text": "A woman speaks with distant quacks of a duck"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/g5l3Bz6lWnc_250.wav", "text": "Insects buzzing as wood lightly shuffles and birds chirp in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ynHdcJ9Oqaw_20.wav", "text": "Loud booming bursts and whooshes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3XuyGJqaXv8_16.wav", "text": "A man speaks over a loudspeaker with a crowd talking in the distance and dogs barking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DNtF_mGzQes_30.wav", "text": "A crowd of children talking as a man speaks over an intercom as a large truck engine runs idle followed by pressurized air releasing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nU-AI3Cmc3M_30.wav", "text": "A pigeon cooing followed by bird wings flapping as wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/K2kIOBeCfuo_140.wav", "text": "An adult male speaks, and a crowd laughs and then cheers and applauds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sqWyxUObwkw_30.wav", "text": "A boat engine starts up and idles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MtK8L8gXRrI_20.wav", "text": "A toilet is flushing followed by a cat meowing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7QN3lwOzfdg_0.wav", "text": "An adult speaks over a telephone and then an adult male speaks, and this pattern repeats three times"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4_DjmCg8Ra8_30.wav", "text": "Artillery guns fire followed by a man yelling as someone whistles while gunshots fire and a group of men shout in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/h5_1pnkl_SY_50.wav", "text": "Middle aged man speaks foreign language while water is trickling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-EQByFLFqig_21.wav", "text": "A man speaks as rain pitter-patters and thunder rumbles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OpiWMltpj44_30.wav", "text": "Ducks quack and water splashes with some animal screeching in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/S0YE96w0YRk_120.wav", "text": "A man speaking followed by a crowd of people laughing then applauding"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pCQEWAFGEjc_30.wav", "text": "A steam engine running on railroad tracks as steam releases and hisses while group of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bQNX7vDalQw_0.wav", "text": "A man speaks followed by heavy sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/GMP8m09j5vk_370.wav", "text": "Birds call while another bird sings"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6Pywt0f_NFY_160.wav", "text": "Water trickling as a machine hums in the background followed by water spraying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/eUecAF626A8_100.wav", "text": "Vehicles hum and vibrate as they rev their engines"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Bvw2gv1fcZw_70.wav", "text": "A person speaks followed by a burping sound and laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SZ6CcXINiiE_0.wav", "text": "A man speaks followed by a loud bursts and then laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oOMtaqvQ3_M_170.wav", "text": "A helicopter flying with wind heavily blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/XplKBvZaHXA_160.wav", "text": "A man talking as a motorbike engine revs and accelerates"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JZloTOdIY_c_30.wav", "text": "Horses neighing and snorting while trotting on grass"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wVi5w_NU6CM_0.wav", "text": "A fire truck siren blasts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9BGLAUSF0sk_30.wav", "text": "An idling engine humming non stop"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bIiiIo20PsY_10.wav", "text": "A man speaking as insects are buzzing and wind is blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-AheI8Epim4_30.wav", "text": "Wind noise followed by a tapping noise while people are talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ir1XTdyt4IY_18.wav", "text": "Released air hissing followed by a popping explosion then a metal ding persists as a person is laughing and a man is talking."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5YzNSjmZ3Wg_20.wav", "text": "A nearby insect buzzes with nearby vibrations"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3MoF8myFs8Y_0.wav", "text": "A strong wind is blowing and constant background waves can be heard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1N_DtRpdAp8_6.wav", "text": "A mid-size engine is idling and vibrating and is revved up one time"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/r2djvq1vc68_20.wav", "text": "A faucet pouring water as water fills a container followed by scrubbing on a plastic surface and water splashing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/d6gu2w19YQo_70.wav", "text": "An baby laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MPLZUg89y5U_90.wav", "text": "Men speak with humming of an engine and some rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/m_NCf-q4Gn0_150.wav", "text": "Plastic thumping as an idle motorcycle engine runs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/E9zN3-C64KE_25.wav", "text": "Scratching noises accompanied by a female voice"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9F3sutgYTvo_30.wav", "text": "A man is speaking and a baby whimpers followed by a woman talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/vfNKduToki4_180.wav", "text": "Truck engines are running, a shrill back-up alarm rings, air brakes engage, and adult males are talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4s2rRnu2PZo_30.wav", "text": "Music plays continuously along with swooshing followed by a man speaking quickly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OxUVcZmeiyI_30.wav", "text": "A clock ticktocks with music"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QvATUKXYFBs_60.wav", "text": "Warning bells ring and a train passes with a honking horn"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Q87LBiwJjTE_250.wav", "text": "Music in the background as a woman speaks and food fries"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/XWw7ZM1c_QA_30.wav", "text": "Wood sanding followed by a woman talking as a clock ticks proceeded by a cuckoo clock alarm cooing and bell ringing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8Zo30kV5aiI_10.wav", "text": "Emergency vehicle siren passing by then fading"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fK4QBQZ6i7w_290.wav", "text": "People are laughing and chuckling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DjKGzOe_COc_30.wav", "text": "A woman is performing a speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/txeXrpoMST4_500.wav", "text": "Water turning on and running continuously"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oN0IcZaHD_8_230.wav", "text": "A man speaks followed by a tools drilling wood"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dJYO3RbBabE_0.wav", "text": "An electronic beep followed by a man talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TtRtURWVYBE_40.wav", "text": "Bells chiming as a clock is ticking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2t82STv2GR8_110.wav", "text": "Several church bells ringing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CwxgQS3SXic_160.wav", "text": "Vibrations from a sewing machine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/VE6Ku0-ucUM_30.wav", "text": "A man speaks followed by a popping noise and laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TWOgvDaDqlU_200.wav", "text": "A power tool drilling followed by a large motor running as wood is being sawed"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QKHpSAAjakY_60.wav", "text": "A motorcycle engine is running, and a man is talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tB8TiiXwKmA_60.wav", "text": "A drill running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dlsiellSFf0_30.wav", "text": "A high frequency motor hums loudly and splashes water"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Dt53UZgyznE_9.wav", "text": "Baby cries and then man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oZaEHkfh5Eg_0.wav", "text": "A vehicle horn honking during a series of electronic beeps"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MVGhC-xB79s_30.wav", "text": "Whistling is ongoing, a child speaks and an adult male speaks, birds coo, and rustling and soft metal clattering are occurring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TaQKhIRwii4_100.wav", "text": "A crowd of people applauding followed by a woman speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EzWEO2WD_MM_70.wav", "text": "High frequency humming followed by a smack"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/twFypUcdgRc_30.wav", "text": "A man yelling as another man is laughing as wind blows heavily into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/W7OJevEgq7w_0.wav", "text": "A dog whimpering and barking while breathing rapidly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7bO0AJI-ihs_30.wav", "text": "Horses trotting on a hard surface as a motor rumbles rapidly for a brief period in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1ed87LLY97k_30.wav", "text": "A fan sound and a tapping noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1QNLMF-Kl_s_30.wav", "text": "A woman is giving a speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tdpiXW68adA_50.wav", "text": "A woman sobbing and speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/sTMKled6Q1M_30.wav", "text": "Whistling with pigeons cooing, wings flapping, and birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5xC4hkAWiao_0.wav", "text": "A large motor vehicle engine is idling, then it labors and creaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-Sz4z0QwEuM_20.wav", "text": "A long, loud burp"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7_smJ8VbfSU_0.wav", "text": "Female speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0yETgW44MZU_20.wav", "text": "A train sounds horn while the engine taps"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dxow2DcTrwk_70.wav", "text": "A man and a woman talking while rain falls, followed by a crash of thunder"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nD1K1Zo0qrM_10.wav", "text": "A series of machine gunfire and two gunshots firing as a jet aircraft flies by followed by soft music playing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0rSETXszQM0_100.wav", "text": "Plastic clacking followed by a motorbike and a motorcycle engine starting up then accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QRtuOWWya30_160.wav", "text": "Metal clacking followed by a man talking as metal rattles while light guitar music plays in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FeHndzYAUkg_170.wav", "text": "A power tool drilling as rock music plays"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dYZSKX7vuRI_300.wav", "text": "A person snoring loudly followed by a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TSnq6n8tElo_0.wav", "text": "A young child talking followed by tapping on glass then metal rattling on a glass surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/T9_ep-3BZDY_300.wav", "text": "A woman speaks followed by crinkling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/inSvboaSRwA_100.wav", "text": "A machine motor running as a man is speaking followed by rapid buzzing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6Nvu6EcpdE8_30.wav", "text": "Wind blows and a man speaks with a quick pop"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AgaiowyYt88_30.wav", "text": "A loud bang followed by silence"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/a3GzZKxUTy8_50.wav", "text": "Birds chirping and then a dog barking and a duck quaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hqPBcvex1VU_18.wav", "text": "An infant crying as a woman mumbles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8ZH_PoK0clI_14.wav", "text": "Footsteps then a woman speaks followed by a door clanging"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8DLcBdC5UrE_13.wav", "text": "Wind followed by splashing of water"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UCy1BEx8jBE_50.wav", "text": "A man speaking as a stream of water sprays and splashes"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ECw5Yf7QoMo_30.wav", "text": "A man speaks while another speaks on top as well followed by clapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zwoqJY03yHE_60.wav", "text": "Women speak and a goat bleats"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/F-47fRplQEc_6.wav", "text": "Different female voices speaking followed by sheep bleating from a distance and voices of young people shortly thereafter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4UPOUGVMlEs_30.wav", "text": "Snorting and breathing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PvWI4p74UOs_30.wav", "text": "Laughing and birds chirping with wind and an engine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Efk5kdn9lR8_21.wav", "text": "Child talking then door crashing and her screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/alaxBd_EEUc_40.wav", "text": "A man speaks briefly followed by a succession of several eructations"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/j1AiqT5oHZc_140.wav", "text": "Beep then male speaking multiple times"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8OTd45_6cvY_30.wav", "text": "Footsteps running on a hard surface followed by liquid spraying then an animal growling as static plays in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/I_vN_BFUr0Y_10.wav", "text": "A train horn blowing several times as a train runs on railroad tracks and a railroad crossing signal rings in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5QZ0NtdoKJ8_200.wav", "text": "Continuous digital beeps with a person speaking and bird chirping in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wBs02amFGXs_30.wav", "text": "A rooster crowing and birds cooing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5I8lmN8rwDM_140.wav", "text": "A power tool sanding"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7P6lcyeDKNI_0.wav", "text": "A bicycle peddling on dirt and gravel followed by a man speaking then laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yhDw7PZje3g_30.wav", "text": "Many insects are buzzing, and an adult male speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nLZeG9LaLgw_0.wav", "text": "Loud humming of a car speeding into the distance with wind blowing and tires squealing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tJhVH3VIrnE_80.wav", "text": "A object is hitting a surface followed by a loud boom"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/v59uHr-B1no_18.wav", "text": "Frogs and crickets chirping during nature ambiance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JQz40TkjymY_30.wav", "text": "Typing on a keyboard is ongoing in a quiet environment"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/R8bHTHnF8j4_100.wav", "text": "A helicopter engine running idle"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rp3CQsWxVgE_0.wav", "text": "White noise followed by a musical horn honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/T32kii824pA_240.wav", "text": "A male speech and metal clanging."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2msevPMQB4M_10.wav", "text": "A power tool vibrating and humming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/59VP93Tzjmg_40.wav", "text": "Distant honks of a train horn with leaves rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/D2Xc_jZllDY_30.wav", "text": "Dog barking with win and man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/a0yXS7PmVR0_30.wav", "text": "Praying with mechanical noise in background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9z2OwpftxUE_10.wav", "text": "Heavy wind and thunder"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PkmpxrsidZM_170.wav", "text": "Applause and female speech and laughter, whistling, and then throat clearing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ptIksg9KEac_30.wav", "text": "A person clapping followed by a dog barking then wood clanking as birds chirp and gusts of wind blow in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8b9z7N25DmU_60.wav", "text": "A bus engine driving in the distance then nearby followed by compressed air releasing while a woman and a child talk in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3XcIVh40pTI_10.wav", "text": "A person is snoring steadily"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oklu5ZJD_2U_10.wav", "text": "Plastic tapping then rattling as a bird chirps"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pHNMcX-9FDs_40.wav", "text": "A dog whimpering constantly then ultimately growls"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LCwSUVuTyvg_330.wav", "text": "A series of doors sliding open as gusts of wind blows and glass clanks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lgwpIImXCWA_180.wav", "text": "A man briefly speaks followed by a sawing sound, and then starts talking again"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/x5AH2gW_8S4_30.wav", "text": "Pigeons cooing with some rattling and screeching"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/p9qRTh4BmSE_30.wav", "text": "Adult males shout in the distance and an adult male shout close by, two adult males speak in the foreground, and gunfire erupts"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ZUmZgPL0ges_20.wav", "text": "Multiple large bells are madly ringing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ndxkSQKxaak_30.wav", "text": "A motorcycle engine is running while a man is talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/u9px4Lwv9XI_30.wav", "text": "Footsteps and a croaking frog"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MvHpNzDpC6Q_50.wav", "text": "Men riding in the car talking then man burps loudly and then the other chuckles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CefFMA3klxk_150.wav", "text": "A horn and then an engine revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rPkCYq4Zjwk_0.wav", "text": "A whoosh effect followed by a woman talking then a man talking proceeded by another man talking over a radio as guitar music plays in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6ukYSXzfEgQ_20.wav", "text": "Metal clacking followed by water flowing and splashing into a container then draining down a pipe"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Jfaj4P3us9M_27.wav", "text": "A phone dial tone playing followed by a plastic switch flipping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tjCNwdOUiGc_120.wav", "text": "A bus engine running followed by a bus horn honking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SQHYl2Kp5ww_10.wav", "text": "Music plays while two men speak each other with dishes clanking and food sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/frOqlk0Wm5Y_30.wav", "text": "A man speaking with intermittent metal scraping."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3ndid3jni7M_40.wav", "text": "A loud horn honking followed by a bell ringing together with clickety-clacking train passing by"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IKnx3hJv1bs_150.wav", "text": "Bad quality audio"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IFRmbxWK8u0_30.wav", "text": "A clock ticking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KJhGuhNHToA_30.wav", "text": "Brief silence followed by wood shuffling on a stone surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6i5eQOpFk_U_170.wav", "text": "Water flowing from a faucet with some light bangs and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LWng-4PDzPM_320.wav", "text": "Rustling followed by whistling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rBUCIK8JRLg_30.wav", "text": "White noise followed by some rustling and splashing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ObWjGBJF_94_350.wav", "text": "Hissing and rattling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9_YfTz8cnFY_40.wav", "text": "Steam hissing followed by a train whistle blowing and a group of people talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3qrVku794u0_12.wav", "text": "A man speaks followed by a beep and a child speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PLHXGDnig4M_3.wav", "text": "A man talking then meowing and hissing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IhvXtS9-IxM_30.wav", "text": "A high pitched engine idling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JdFmMw0zyKA_0.wav", "text": "A kid speaking followed by a seal making noises then water splashing as a kid and woman are laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1HCuBnPLMqQ_380.wav", "text": "A liquid makes a trickling noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NlKlRKz8OKI_90.wav", "text": "A woman talking then laughing followed by plastic tapping and birds chirping in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xUWSHYoslPQ_90.wav", "text": "A man is getting a glass and setting in down on a hard surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6pssFJ0m-kU_60.wav", "text": "Leaves rustle in the wind as small birds chirp"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0NGSrwioYjA_30.wav", "text": "A man speaking followed by a goat baaing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/RfGapDlAYoQ_30.wav", "text": "Musical whistling with wind blowing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/dP5DbAzTl5M_30.wav", "text": "A motorboat engine running and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/glAeihz0NAM_130.wav", "text": "Kids laughing then talking followed by a young man talking as wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CMUuelJFJ7Q_30.wav", "text": "A bell ringing followed by a clock ticking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QHfyKaOHSz4_0.wav", "text": "A fly buzzing and a frog croaking with musical sounds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bLZFtoWXYTA_0.wav", "text": "A cup is filled from a faucet"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jYPU6aSDo88_50.wav", "text": "An airplane engine and then light talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/o7jW6Suyfbs_80.wav", "text": "A man talks in the distance followed by several vehicles driving and accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/i1u_2eZYYlE_0.wav", "text": "An idle vehicle engine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_ezm-TpKj1w_30.wav", "text": "Humming of a revving idling engine with people speaking in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2a6GNu6uCDE_30.wav", "text": "A woman delivers a speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Z1Cyj4N05lk_140.wav", "text": "Whistling with small rustling and groaning"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AWGnTI0e2Fs_300.wav", "text": "Helicopter propellers spinning"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/N_s9F4CI_98_380.wav", "text": "Pigeons cooing followed by bird wings flapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AUJPx81qKtY_0.wav", "text": "A man talks and insects buzz"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pO8kbg9IJnc_120.wav", "text": "Some light clicks followed by a man speaking and water splashing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yfYNPWs7mWY_380.wav", "text": "A man speaking, followed by a door shutting, and then the man speaks some more"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/niwgMbB6tpQ_10.wav", "text": "Plastic clicking and clacking followed by a woman snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8DQfjqPCTI8_450.wav", "text": "Swarm of bees buzzing and two men speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lfO471Rn61k_230.wav", "text": "Steam escapes with a hissing noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hVUmQfBIYe8_100.wav", "text": "A man speaks followed by a door opening"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/SCow4mpBsGY_0.wav", "text": "A person is snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ot0bN_hz2ec_20.wav", "text": "A train horn honks with a train racing by"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3rna9zo5ZOs_30.wav", "text": "A man is giving a speech and a crowd cheers"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/NeWW30WZjPc_10.wav", "text": "A dog barks and rustles with some clicking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Fc9pG54DDJM_30.wav", "text": "A toilet flushing followed by paper crinkling and a door slamming in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/O90Qy2xG6oA_30.wav", "text": "Crying and screaming with some coughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3xDZ-kdGE3o_10.wav", "text": "Metal clacking followed by a door closing as water trickles then a toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Xz56Q2Q5j5c_570.wav", "text": "Wood being scraped along with mechanical sounds"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KSHpYhuTotY_210.wav", "text": "A man speaking as footsteps walk on grass while insects are buzzing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/VOXl8iR-HnI_240.wav", "text": "Quiet speech and then and airplane flying away"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CO6-i8NLbeo_0.wav", "text": "A man talking followed by a goat baaing then a metal gate sliding shut as ducks quack and wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-JP1GqPEKtw_130.wav", "text": "A man talking as a power tool motor runs followed by the buzzing of an electric sanding machine"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EcihYbSlyck_30.wav", "text": "Vehicles are moving, horns are sounding, the door alarm to a bus is beeping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/14izd_i3ryE_30.wav", "text": "A man speaks as an engine runs and wind rushes by, another man speaks over a loudspeaker"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wOFBldBFRNk_20.wav", "text": "Church bells ringing as rain pours and wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FfUqv0Vv3ME_20.wav", "text": "A horse walks while an adult male speaks, an adult female speaks briefly, and roosters are crowing in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/naPgJvWTIY4_80.wav", "text": "A motorbike running idle then revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tNxfdAd14qE_30.wav", "text": "Mechanical rotation and then a loud click occurs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/os_2U4xqTqw_100.wav", "text": "Heavy footsteps walking on a hard surface followed by a man grunting then two explosions"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7RMpCCkQks0_20.wav", "text": "An industrial motor is humming, and ripping and tearing are occurring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Jp64Whpr3BA_50.wav", "text": "Rustling followed by a sewing machine running."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tmLAXm1WlnE_50.wav", "text": "A woman speaks with nearby buzzing of insects"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/22L_3pBa1AI_18.wav", "text": "A car is shifting gears"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ne2DpKCIr4Y_190.wav", "text": "Ocean waves crashing as wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ls2vrr9TamU_120.wav", "text": "Drilling and crackling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ESjMIqrvRj4_30.wav", "text": "Rain pitter-patters as thunder cracks and the wind blows"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OUUckswAaNI_100.wav", "text": "Male speaking and birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AMQei29haCw_20.wav", "text": "Buzzing and humming of a motor with a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9U8COLzEegs_420.wav", "text": "Water running softly in the background followed by several shrill beeps and a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8ycflE3dIHw_160.wav", "text": "Train passing followed by short honk"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0ury8KHQdL4_480.wav", "text": "An adult male is speaking and typing on a keyboard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DAN1t9ukkg0_30.wav", "text": "A person is typing keyboard with some mouse clicking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/TQr9v-PQOc4_0.wav", "text": "A person sneezes and coughs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JTHMXLC9YRs_30.wav", "text": "Breathing and ducks quaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WUpeplQr3A4_300.wav", "text": "A man shrilling followed by a power tool drilling and a man screaming as liquid splatters"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/V8A0VRGdgwM_10.wav", "text": "Pebbles and rocks bouncing off of a wooden surface as footsteps walk on gravel and dirt followed by a goat baaing as a muffled animal moan in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mlnUJH4BQnk_110.wav", "text": "Female speech and scratching"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kEP-BwMarf8_90.wav", "text": "A wrapper crinkling and a woman's soft voice"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/F7QtqKtllK0_50.wav", "text": "A person snores"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1L_OyngNZMA_30.wav", "text": "Two men talking on a microphone as a crowd of people laugh and applaud"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/C5kmOK_l4jc_0.wav", "text": "A young girl talking as an infant is laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hDMHIDJdfDA_40.wav", "text": "Continuous snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Z3wDry8nnJs_60.wav", "text": "Water splashes followed by a girl speaking then scraping followed by spitting"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/z1ax0QPpd14_170.wav", "text": "A large duck quacks and small ducks chirp"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0Dt-pH0pW-Y_20.wav", "text": "A motorcycle running and male speech and then laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/eu5bq0A3XVQ_250.wav", "text": "A male voice talks after doing some burps"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BA-lFjpzad4_30.wav", "text": "Motor vehicle traffic is ongoing, and a motorcycle engine accelerates and then decelerates in the foreground"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0yxEvdnimGg_30.wav", "text": "Dogs are barking with wind noise while people are speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/e2rScj9UyMs_30.wav", "text": "Pigeons cooing and bird wings flapping as footsteps shuffle on paper"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CvNAwby6Xos_220.wav", "text": "A sewing machine operating several times followed by gears cranking as a man is talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xBZnvfniA1c_30.wav", "text": "A man speaks, and a child speaks followed by laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/L6rnV0oNIII_8.wav", "text": "Laser followed by a gun shot"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/H-vTZh81qAU_30.wav", "text": "Clicking followed by humming noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FR7BDRhMATo_310.wav", "text": "A train running on railroad tracks as a train horn whistle blows several times while railroad crossing warning signals are ringing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xqtrbqDlz28_20.wav", "text": "A sneeze occurs followed by sniffling and the start of another sneeze, while a baby coos in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JhGp7HmRQxg_0.wav", "text": "A metal clack followed by a horse neighing as gusts of wind blow and birds chirp in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/taYKM1OSTwE_120.wav", "text": "Sheep baaing as gusts of wind blows followed by a person chuckling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zoxFl3pddMg_2.wav", "text": "A frog chirping with distant speaking of a person"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zF3xXn6NTyU_260.wav", "text": "Water falling and a woman talks, coughs, then talks again while a man speaks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zq00Oe1ecpE_140.wav", "text": "A bus engine slowing down then accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/OVQMFBeCHq0_13.wav", "text": "Low frequency emergency siren going off followed by the siren slowing briefly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-oy0BkpMGAk_30.wav", "text": "Revving of an engine and a skidding sound"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Axd__X2rixk_30.wav", "text": "Clip clops of a horse running by"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/oNHCc_izsDE_10.wav", "text": "Water splashing and a baby laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Or7umk40TZA_6.wav", "text": "An engine is idling with vibrating noises"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9hxFqltp3xw_30.wav", "text": "A woman speaking followed by spraying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fsBR7e_X_0Y_40.wav", "text": "An infant yelling as a young boy talks while a hard surface is slapped several times"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0On6-JiVwRs_100.wav", "text": "Cat meowing sadly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QOmV7O9mFwg_30.wav", "text": "Laughing followed by a child talking and more laughing with many children talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/kVYNXZd0MMY_30.wav", "text": "An idle vehicle engine running followed by bird cawing in the background and a plastic camera click"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3ue0gJM0THk_30.wav", "text": "Aircraft engine revs and creates propeller which"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jjfUaMQaG1A_310.wav", "text": "A man talking followed by plastic clacking then a power tool drilling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Z-SIyOChVh8_0.wav", "text": "Rain and thunder"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/woadpeAGHUQ_70.wav", "text": "An emergency siren blaring steadily"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/096oTVzc5Gs_30.wav", "text": "An adult female speaks and an animal growls playfully"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/3VHpLxtd498_30.wav", "text": "Humming with coos of pigeons and brief distant speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/vaujJ7msKfc_150.wav", "text": "Wind blowing with distant traffic passing and sirens ringing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DlWd7Wmdi1E_150.wav", "text": "A man talks while several animals make noises in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2ceUOv8A3FE_20.wav", "text": "A train horn blowing multiple times as a train runs on railroad tracks while a man and a young kid talk in the background alongside birds cooing in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lmPMhs-9IYE_0.wav", "text": "A vehicle engine revving as a crowd of people talk and whistle in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/q3SEOW2m4WY_20.wav", "text": "A train horn honks as a train speeds past"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/D4s5aHrsBgs_80.wav", "text": "A person whistles to music"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Pb6MqpdX5Jw_40.wav", "text": "A horse gallops then trot on grass as gusts of wind blow and thunderclaps in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0_K6OKtoBBU_30.wav", "text": "A cat is meowing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1_z6NcidGzM_30.wav", "text": "Various speech and a whistle blown and water splashing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LvMA1Wcgu3w_30.wav", "text": "A man talking followed by plastic creaking and clacking as frogs croak and crickets chirp"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rbO727iF03I_30.wav", "text": "A loud belch followed by male human voices"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ZBtgrP4vU_w_50.wav", "text": "Food and oil sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tfOIhQpYYe8_60.wav", "text": "A helicopter flying as a man talks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/2ABngPM3raQ_30.wav", "text": "A man talking as music is playing followed by a frog croaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jlwe9jtu5Gw_50.wav", "text": "A couple of whistling sounds one after another"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7fmOlUlwoNg_20.wav", "text": "A machine is making clicking sound as people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-CcGuq0yoKo_30.wav", "text": "A woman speaking on a microphone as a kid yells in the distance with periodic audio static clicks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4ftDFi4684Y_30.wav", "text": "Several slight bangs occur during a constant ticktock"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/AtkD-3GjXMw_40.wav", "text": "Singing and music with faint gunshots"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/e4ph6bIC5zc_170.wav", "text": "Distant murmuring and distant engine revving"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/y93cZqNCtks_470.wav", "text": "A man talking as footsteps run on concrete followed by a series of gunshots firing a gun cocking while another man talks in the background before a dog growls"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KtinboYbmHQ_50.wav", "text": "Tires squeal and an engine hums as a vehicle approaches"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/zBXoaQ1GVlc_380.wav", "text": "Several children speaking and screaming with a woman speaking briefly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/mJ6ZO3xEcgw_50.wav", "text": "A woman sneezing repeatedly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9ZZHvwaH-CU_160.wav", "text": "Multiple gun shots woman screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/uJzAf4PaExI_180.wav", "text": "An engine idling consistently"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/aZAXO2WZn84_30.wav", "text": "A train running on railroad tracks followed by a lawn mower engine running then a steam engine whistle blowing as a crowd of people are talking in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/tpm5IOD5d4o_30.wav", "text": "An adult female speaks and claps, a bird whistles in the foreground, another bird chirps in the background, and then a human whistles a tune"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-NrFeH-kBSM_0.wav", "text": "Clicks of weapons being loaded followed by a gunshot and a clink then a man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_YS5uKWoB6g_30.wav", "text": "A baby cries as adults speak with clicking and banging of a car door"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6BJ455B1aAs_0.wav", "text": "A missile launching followed by an explosion and metal screeching as a motor hums in the background."}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/K03ydb1uaoQ_40.wav", "text": "Static sounds are vibrating along with snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/gW7s3YAthpI_430.wav", "text": "A person gulping followed by glass tapping then liquid shaking in a container proceeded by liquid pouring before plastic thumps on paper"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/bgbnu5YKTDg_30.wav", "text": "A man is speaking followed by gunfire"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fx4r_KuW6No_50.wav", "text": "Adult female is speaking and a young child is crying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4fz0-Kx2oNs_250.wav", "text": "Food and oil sizzling followed by a person snapping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4CAMv5nlr-0_30.wav", "text": "A man speaking followed by a crowd of people applauding and cheering"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/hzn_wGlzGpU_90.wav", "text": "An aircraft engine accelerating followed by an object clanking against fiberglass as an emergency siren sounds in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9b6RqajfAmw_30.wav", "text": "Birds chirping together with rustling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xpZna_FwDhI_70.wav", "text": "A click is followed by a woman speaking and then a sewing machine stitching"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Tdl9SmBbRnA_50.wav", "text": "A man is talking as a motor vehicle engine travels"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/eYbFtxZmKL4_110.wav", "text": "Clanking and rustling with people faintly speaking and horse clip-clopping in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/KOBkbROPv4c_0.wav", "text": "Humming of an engine with squealing tires"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Q0anPAIkfBE_60.wav", "text": "A baby sobs and then a woman speaks followed a woman laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BGEMgl1xjac_120.wav", "text": "Various insects and bugs are chirping with a rodent breathing sound in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Wq4OD3olO2w_90.wav", "text": "A man talks followed by laughter, and a woman and children communicate"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ajheseWZmmU_70.wav", "text": "A cat meowing as a person makes kissing noises followed by a guy chuckling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BlbGXalLNVU_70.wav", "text": "Surf comes ashore as a man speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/86dNVnTwH6U_40.wav", "text": "Hissing followed by sewing machine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ek9Fsmm3xqk_260.wav", "text": "Splashing with wind blowing hard"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pPLisQ_QXxw_24.wav", "text": "An engine running and a horn"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/fBYDJWChe5c_20.wav", "text": "A person snoring"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7upINC4seBw_30.wav", "text": "A man talking as an idle motorboat engine runs and water trickles while wind is blowing into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Xi6V0LGvqoo_0.wav", "text": "A dog barking followed by a dog growling as footsteps squeak on a hard surface floor and followed by series of metal clicks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EYTz1LPDHsc_30.wav", "text": "A car door opening as a crow caws and birds chirp followed by a vehicle driving by in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/cz0FSQDVBMw_90.wav", "text": "Hissing and chugging with brief muffled speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_AcJVyToQUQ_60.wav", "text": "Laughter and a female speaking and then a male vocalizing a sneezing sound"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nlC4UI4hZ60_10.wav", "text": "An engine being turned over"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/inQOrxc_oZo_30.wav", "text": "Hog snorting with male screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/EY4p0_NJVQs_30.wav", "text": "A woman is performing a speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5G6b_QWL3nY_60.wav", "text": "A woman is speaking while food is frying and sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/A0E_UiD-fR4_0.wav", "text": "Digital beeping followed by rapid vibrations and humming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wbPmnxCLoRQ_40.wav", "text": "Several bells ringing rapidly and then a single bell"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/9ucb5HYO8ps_20.wav", "text": "A burp followed by laughter and speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/z4MeV9IGVo0_30.wav", "text": "Humming from loud engines with men speaking over the radio"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WWkhzcmx3VE_320.wav", "text": "A duck quacking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/R91bUbtKrRs_30.wav", "text": "Two women are talking, and a baby cries"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/nmLMLgWPmFM_240.wav", "text": "A motorboat engine running as water splashes and a man shouts followed by birds chirping in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/WU3qB7gf6ao_30.wav", "text": "An electronic beep and several faints ticks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/W4GEwnXc9tQ_30.wav", "text": "While music plays in the background, a frog croaks, young females speak and laugh, and a young male speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/BOB65Nd0pXo_180.wav", "text": "A helicopter engine operating while wind blows heavily into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/CM49C3RkzV8_30.wav", "text": "A woman delivers a speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/cNARVD02-tw_530.wav", "text": "A man and an elderly woman talking as plastic clacks followed by a faucet pouring water as water runs down a drain"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/63KW_EQ72yU_30.wav", "text": "A loud explosion as gusts of wind blow"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ITlqMkR5alY_0.wav", "text": "Wind blows hard followed by screaming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/xQDq3A4Zfbo_30.wav", "text": "Water splashes and wind blows"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/YQGW5AwDOIo_560.wav", "text": "Motor vehicles are driving with loud engines and a person whistles"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/XPebkNzsnRI_10.wav", "text": "Whistling followed by a child giggling and then Moe whistling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ff8bCCJfVX4_10.wav", "text": "An air gun firing rapidly followed by plastic rattling and a young boy talking while footsteps walk over grass"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/I_8KqxP5xOA_50.wav", "text": "A sprayer is spraying surfaces"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/11SEBDuoqSk_50.wav", "text": "An engine chugging loudly together with gunfire"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/DzKjogSVOLM_21.wav", "text": "Ducks quacking and birds chirping followed by a rooster crowing as a crowd of people talk"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/IdBDl9Wr51A_10.wav", "text": "Synthesized laser effects and explosions as a man talks followed by a deeper voiced man talking then woman laughing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1slvoNgzBLE_160.wav", "text": "An electronic signal followed by compressed air releasing then an electronic bell playing as a train runs on tracks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/H7rd9bZtbgc_330.wav", "text": "Bells ringing repeatedly"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/szkiW0GXEOI_100.wav", "text": "Whistling with birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/atmDP_fmK_8_0.wav", "text": "A low loud hum gets lower followed by an explosion"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/ArHiac57pVk_30.wav", "text": "A male speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/up2PpjTzyyc_30.wav", "text": "Electronic beeping followed by a cat singing then meowing as paper shuffles and a man talks with music playing in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/B3O476LeuXY_30.wav", "text": "An idle motorbike engine running"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JnSwRonB9wI_30.wav", "text": "A muffled motorboat engine running as a group of people are laughing and screaming while water splashes and sprays"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rN2rpLV3brs_120.wav", "text": "Dishes clanking as a man is talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4pf-PIymDhU_20.wav", "text": "A jackhammer drilling and vibrating continuously"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1IoHRTUp86c_70.wav", "text": "An aircraft engine humming as a woman speaks over an intercom followed by an aircraft engine accelerating"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/-R69Fa-mCaY_20.wav", "text": "A chainsaw cutting as wood cracks and creaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/MjSegUnQXr4_2.wav", "text": "Pigeons cooing and bird wings flapping during light scrapping on a metal surface while a construction machine operates in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/7XXSOzDQ2z0_70.wav", "text": "A mid-size motor vehicle engine is idling and clacking somewhat, followed by mid-size motor vehicles accelerating quickly with a brief tire squeal"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UQtBt6CQpwg_30.wav", "text": "A man speaks, and a sewing machine sews material"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/8ipe6b1LwHQ_30.wav", "text": "A glass link occurs, a thump, and then clattering"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/qZEIs6tS5vk_0.wav", "text": "A vehicle engine revving followed by tires skidding as a group of people talk in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/XamQAY_WXRY_140.wav", "text": "A man talking as water splashes and gurgles and a motor engine hums in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PMMkPq5jJXY_0.wav", "text": "A person burps as people laugh"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/0_ogYGDGDco_60.wav", "text": "Water is splashing, people are talking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/K_Vre_-4KqU_60.wav", "text": "An aircraft engine runs and vibrates, metal spinning and grinding occur, and the engine accelerates and fades into the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/r2KhpX_QgXA_230.wav", "text": "A man speaks as a machine hums quietly and then increases in volume"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pgq2KPX5_SA_70.wav", "text": "Some crinkling of cellophane"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/6aWnK1GyeJY_30.wav", "text": "A baby is crying and a person sneezes then another person speaks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/JmWaRt8-u0s_480.wav", "text": "A motorcycle engine running idle while two men communicate with each other"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1nUOGZgSzZo_30.wav", "text": "Water is splashing and a man laughs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/t1hj7se76wQ_30.wav", "text": "Typing on a typewriter is ongoing with clicking and clacking, a bell rings, and a clack occurs with the carriage return"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/jid4t-FzUn0_20.wav", "text": "A man speaks followed by a goat bleat"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/yRoKi7rhSRo_20.wav", "text": "A toilet flushes and water drains"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/5t6tSW0yT40_410.wav", "text": "A motor hums softly followed by spraying"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/_GI7meqlYZk_30.wav", "text": "A cat screaming then meowing as a young boy is talking during a series of electronic buzzing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/elztUCeNQvQ_10.wav", "text": "A train horn honking followed by a subway train running on railroad tracks"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/PMMdAKZxI_I_40.wav", "text": "Burping and laughing by girls with distant speaking of a television"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/lrKGCtSsAkA_50.wav", "text": "A toy helicopter engine flying as wind blows into a microphone"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Gkb4f6yodPE_70.wav", "text": "Sustained industrial sawing noise"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/y-RSojxgkDo_0.wav", "text": "A male speech and a bird chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LF6x7B0Ppvo_20.wav", "text": "A racing car is driving fast"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/puZL08fzpXk_4.wav", "text": "Gunshots followed by people speaking and explosions"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LP_DzNUkAKY_490.wav", "text": "People speaking faintly with some humming and a hydraulic hiss"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rtgVoZCcBw8_0.wav", "text": "Goats are bleating with other natural outdoor noises"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/FDwK7T1JO_0_160.wav", "text": "A male voice and then drilling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Ni3dIj90Oa4_10.wav", "text": "Explosion followed by man speaking a short bit of music at the end"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/4YodC6RnplI_70.wav", "text": "A female laughs, snoring occurs, and an adult male speaks in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rvtA7c1I4xo_0.wav", "text": "Whooshing of wind followed by a man speaking with sneakers squeaking on hardwood"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/wFiCblfZ-vg_30.wav", "text": "A man speaking followed by another man speaking in the background during audio static"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/L8dA-2Lu2hY_20.wav", "text": "A person whistling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/b1PXsfgQw5w_0.wav", "text": "A group of people laughing followed by a man talking then more laughter"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/rUq4w4EUSWA_0.wav", "text": "An electronic motor buzzing and paper tearing followed by a bell chiming and a plastic click then a toilet flushing as paper crinkles in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Agh2EKINlSw_30.wav", "text": "Hard rain falls on a metal surface, then three adult males speak while soft rhythmic whooshing occurs"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/HqndxoujCYI_30.wav", "text": "A bell chiming as a clock ticks and a man talks through a television speaker in the background followed by a muffled bell chiming"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/1FNJbN-eHY4_30.wav", "text": "Burping and girls laughing with many voices in the background"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/473wBEwC35M_30.wav", "text": "A man speaking as a vehicle horn honks and a man speaks in the distance"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/f2fSxfvmkZQ_310.wav", "text": "A man talking as metal thumps followed by plastic rustling then a power tool drilling while a metal pot clangs against a hard surface"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/QTSKjweEWew_170.wav", "text": "Wind blowing followed by a man speaking"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/Jon_DEFqsfM_30.wav", "text": "Wind blowing with ducks quacking and birds chirping"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/UmNrhFKpWIY_21.wav", "text": "A car engine revs and then shuts off"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/y3-M1sonh3M_0.wav", "text": "Clinking followed by a toilet flushing"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/g6CY7qvu81k_0.wav", "text": "Music followed by male speech and distant sizzling"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/pI_kPedctoo_10.wav", "text": "An engine running and being revved and some muffled speech"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/a_Rjlu50TfA_0.wav", "text": "A series of sharp, squeaky snoring noises"}
+{"dataset_name": "AudioCaps", "audio_path": "/data/audiocaps/data/LKhokVsJhN0_210.wav", "text": "A flock of sheep baaing"}