Datasets:
audio audioduration (s) 496 540 | genre stringclasses 6
values | title stringclasses 6
values | speaker stringclasses 5
values | category_path stringclasses 6
values | track_id stringclasses 6
values | source_url stringclasses 6
values | clip_offset_sec int64 0 480 | duration_sec float64 496 540 | text_gemini stringclasses 6
values | transcription_model stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|
torah_shiur | ืคืจืฉืช ืืืจืื | ืืจื"ื ืจ' ืืืืืื ืจืืืื ืืืืืืฉ ืฉืืื"ื | cat/230 (ืชืืจื ืจืืืจืืง) | 161840 | https://cdn.yiddish24.com/DEVUREM_230_1784309278982_333_0_260717014_132815621.mp4 | 180 | 540.024 | ืืืื ืืื ื'ืืื ืื ืืขืืืื ืืื ื, ื'ืืขื ื ืืขืฉื ืก'ืืื ื ืืืงื... ื ืืืจื'ืืืืข ืืืจ. ืืฉื ืจืื ื ืืขืจ ืจืื ืฉื ืืฉืจืื, ื'ืืืจืฃ ืืื ืืขืจื ืขื ืก'ืืื ื ืืืกืจ ืคืืจ ื ืืขื ืืฉ. ืืขืืขืจ ืืขื ืืฉ ืืืจืฃ ืืขื ืืขื ืืืกืจ ืืื ืืขืจื ืขื ื ืืืกืจ ืืฉืื. ืื ื'ืืขื ืื ืืขืฉืืืช, ืืื ืืขืื ืขื ืื ืก'ืืืืืื ื ืืืกืจ ืฆื ืืืจ. ืจืฉ"ื ืืืื ืื, ืจืฉ"ื ืืจืขื ืื ื ืจืฉ"ื ืืืืก ืขืก ืืื ื ืคืื ืก'ืืื ื ื... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) | |
news | ืืงืืืขืืข ื ืืืขืก / ืกื ืืื ืคืื ืื ื ืืืขืก (3 bulletins concatenated) | null | cat/59 (ื ืืืขืก โ ืืืืขืืืื ืข ื ืืืขืก) | 162705;162636;162550 | 0 | 496.44 | ืืงืืืขืืข ื ืืืขืก, ืืื ืืื ืคืจืฉืช ืขืงื.
ืก'ืฉืืขืื ืืื ืืจืืืก ืื ืื ืจืืืืืข ืืืืืขืจื ืข ืืืงืกืขืก ืืื ืงืื ืกืืจืืงืฉื ืืจืืขื ืืื ืืขืจ ืื ืืจืขืื ืืขืืื ืืืขืจ ืืืคืขืจ ืืื ื ืืฉื ืงืืื ืืืืขืืืืข ืืื. ืื ืืืฉื (OSHA) ืืืขื ืืืจ ืืื ืืขืฆืื ืก ืืืจืืืขืคืืจื ื ืืืืืืืข ืืืืืขืจืืืืข ืืื ืกืคืขืงืฉื ืก ืืืืฃ ืงืื ืกืืจืืงืฉื ืืจืืขื ืืืืขืืื ืืื ืืื ืืขืื, ืืื ืืจืืืกืืขืืขืื ืงื ืกืืช ืคืื ืืื ืืขืจืืขืจ ืืื... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) | ||
interview | ืืืืขืืืื ืข ืืื ืืขืจืืืื: ืืจื ืืื ืคืื ืืขืจ - ืขืืขืืื ืืืกืืจืืืืจืืขืจ | ืืจื ืืื ืคืื ืืขืจ | cat/38 (ืืื ืืขืจืืืืืก โ ืืืืขืืืื ืข ืืื ืืขืจืืืืืก) | 160984 | 480 | 540.144 | ื'ืืื ื ืกืื ื ืื ืคืื ืืืืข'ืก ืฆื, ืื ืื ืืื ืฆืข ืืฉืคืื ืืื ืืืขื ืืืข ืืืื ืืืืขืงืืขืฉืืขืื ืืกืืืืฉืข ืืืื ืืื ืืืื ืขืก. ืืื ืืืืขืงืฆืืฉืืขืื ืืจืื ืืฉื ืชืืจื ืืืกื ืก'ืืื ืืขืืืขื ื ืกืื, ื ืงืืืืขืจ ืืขื ืืฉ ืืื ื... ืืื ื ืืืืขืจ ืืขื ืืฉ.
ืืื ื ื'ืืื ืืื ืฉืืื ืืขืืขืจื ื ืืื ืืืืืืืืืืฉ ืืืก ืงืื ื. ืก'ืืขืืืขื ืฆืืืื ืืืจืื ืืขืืืืืก, ืืืขืจ ืก'ืืขืืืขื ืืืขืจ... ืืื ืืืื... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) | ||
monologue_podcast | ืงืื ืืืข - ื' ืืืจืื ืค"ื | ืืฉื ื ืืื ืงืจืืืก | cat/206 (ืคืืจืฉืืืขื ืก โ ืงืื ืืืข) | 161828 | 90 | 540.029388 | ืขืก ืืื ื ืืืืขืจ ืืืืืืื ืืขืืข ืืืืก ื'ืืืื ืจืขืื. ื'ืฉืคืืจื ืืื ืืื ืืื ื ืืืื ืืื ืก'ืืขื ืขื ื ืืฉื ืงืืื ืกืืกืืืขืื ืื ืืืืฃ ืื ืืืขืื. ื'ืืืื ืืืขื ืืื ืืื ืก'ืืขื ืขื ืืื ืืจืืืกื, ืืืื... ื'ืงืขื ืืื ื ืืฉื ืืืื ืฉืืืกื. ืืืขื ื ืืขื ืืฉ ืืืื ืืืืจื ื ืฆืจื ืฆื ื ืืืืื, ืฉืคืืจื ืืื ื ืืฉื ืคืืจืฉืืื ืขื. ืื, ืก'ืืื ืก'ืืื ื ืืืจื'ืืืืข, ื ืืฉื ืืืืข ืืจืืฉื ืืืืก ืืขืืขืจ ืฉืค... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) | ||
general_podcast | ืืฉืืงืืืืข ืืจืขืงืืขื - ืืืฆ"ืฉ ืืืชืื ื ืค"ื | ืืืฉืข ืืืืืก | cat/288 (ืคืืจืฉืืืขื ืก โ ืืฉืืงืืืืข ืืจืขืงืืขื) | 162485 | 60 | 540.024 | ืืืื ืืืจ ืืืื ืฆื ืืื ืืืืขืจ ืืืฉืื... ืก'ืื ืืื ืืื ืื ื ืืขื ืืฉ ืงืขื ืืืงืืืขื ืืื ื... ืคืื ืืขืืืืกืข ืคืืื ืืืขืจ ืืื ืืขืืืืกืข ืคืขืืขืจ ืงืขื ืืขื ืืืงืืืขื ืคืื ื ืืขื ืื ื ืกืื ื ืืจืขืกืขืจืข ืืืจืขื ืกืืคืืจื ืืื ื ืืขื ืืฉ. ืก'ืื ืืืืก ืืืื ืื ืื ืจืืื ืืื ืคืืขืื ืืืืื ื ืืขื ืื ืงืขื ืืื ื ืืฉื ืจืขืื ืืื ืฆืืจืืง, ืืื ืขืจ ืืขืจื ืืื ืืืืก ืืื ืขืจ ืืขื ื'ื ืืื ื ืืฉื ืฆืืจืืง... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) | ||
vlog | ืขืืืฉ ืืืืง ืงืืืคืื ื (ืืืื ืคืขืจืืืืืืืฉ ืฉืื) | ืืืื ืคืขืจืืืืืืืฉ | cat/242 (ืืืืืขื โ ืืืื ืคืขืจืืืืืืืฉ ืฉืื) | 162532 | https://videos.yiddish24.com/20260725102035-162532.mp4 | 60 | 540.024 | ืืขืจ ืืื ืืื ืขืงืกืืืืขื. ืืฉื ืืื ืขืงืกืืืืขื. ืืึธืืืจ ืืืื, ืืึธืืืจ ืืืื, ืืึธืืืจ ืืื ืืืืึทืื, ืืึธืืืจ ืืื ืืืืึทืื.
ืึธืืจืืื, ืึธืืจืืื, ืืึธ ืืขื ืขื ืืืจ. ืืืื ืืื ืื, ืืืื ืืื ืื, ืืึธืืืจ ืืื ืืึธ ืึธืคึผืฉืืขืื ืึท ืืื ืื ืคืึทืจ ืืืจ ืืืืขื ืึทืจืืืก, ืคืึทืจ ืืืจ ืืืืขื ืึทืจืืืก ืืฉื. ืึธื, ืืืืึทื ืืึธ. ืคึผืื ืงื ืืึธ. ืืืื-ืืืื ืืื ืงืขืจ (YY Bunker) ืืึธืงื ืคืึทืืจืืง... | google/gemini-3.1-pro-preview (via Vercel AI Gateway) |
Yiddish ASR Benchmark (Multi-Genre)
A small, genre-diverse Yiddish ASR benchmark seed set: 6 clips (~53 minutes total), one per
genre, each with a Gemini-generated verbatim reference transcript. Intended to complement the
existing single-speaker Yiddish24 collections (Kohn-AI/yiddish24-audio,
Kohn-AI/yiddish24-wav, ABE101/yiddish24-dual-asr) with broader genre coverage for
benchmarking rather than training.
All source audio is from yiddish24.com.
| field | meaning |
|---|---|
file_name |
audio (mp3) under data/ |
genre |
one of torah_shiur, news, interview, monologue_podcast, general_podcast, vlog |
title |
episode/segment title |
speaker |
presenter/host name, where applicable |
category_path |
yiddish24.com category the source came from |
track_id |
yiddish24 post id(s); news clip concatenates 3 bulletins, ids ;-separated |
source_url |
original cloudfront/cdn url(s); ;-separated for the concatenated news clip |
clip_offset_sec |
start offset (seconds) into the original recording where this clip begins |
duration_sec |
clip duration |
text_gemini |
reference transcript (Yiddish, Hebrew script), verbatim, unclear spans marked [?] |
transcription_model |
model used to produce text_gemini |
Genres
| genre | title | source |
|---|---|---|
torah_shiur |
ืคืจืฉืช ืืืจืื | ืืจื"ื ืจ' ืืืืืื ืจืืืื ืืืืืืฉ ืฉืืื"ื |
news |
3 concatenated news bulletins | Kol Mevaser news desk |
interview |
ืืจื ืืื ืคืื ืืขืจ - ืขืืขืืื ืืืกืืจืืืืจืืขืจ | ืืืืขืืืื ืข ืืื ืืขืจืืืืืก |
monologue_podcast |
ืงืื ืืืข | ืืฉื ื ืืื ืงืจืืืก |
general_podcast |
ืืฉืืงืืืืข ืืจืขืงืืขื | ืืืฉืข ืืืืืก |
vlog |
ืขืืืฉ ืืืืง ืงืืืคืื ื | ืืืื ืคืขืจืืืืืืืฉ ืฉืื |
6 clips ยท ~53 min.
Reference transcript method
text_gemini was produced by google/gemini-3.1-pro-preview, called via the Vercel AI Gateway
chat completions endpoint with the audio file attached as a file content part
(media_type: audio/mpeg) and a verbatim-transcription prompt. These are a first-pass gold
standard โ worth a human spot-check before treating as strict ground truth, especially for
proper nouns and numbers. Unclear audio is marked [?] in the transcript.
from datasets import load_dataset
ds = load_dataset("Kohn-AI/yiddish-asr-benchmark")
- Downloads last month
- 46