datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Whisper-Hallucination
Whisper Hallucination and Repetition Probes
This is a benchmark. Every evaluation config is test — do not fine-tune on it.
lexicon_synth is the exception: synthetic training material with its own train/test
split, and not one of the eight benchmark arms.
To build training data, exclude the items in
benchmark/exclusions.json
(546 FMA tracks, 1,168 FSD50K ids, 2,620 LibriSpeech utterances, the Malay stems). The
benchmark draws FSD50K eval and FMA shards 0–1, so training can use… See the full description on the dataset page: https://huggingface.co/datasets/Scicom-intl/Whisper-Hallucination.arabic_speech_corpusThis Speech corpus has been developed as part of PhD work carried out by Nawar Halabi at the University of Southampton.
The corpus was recorded in south Levantine Arabic
(Damascian accent) using a professional studio. Synthesized speech as an output using this corpus has produced a high quality, natural voice.
Note that in order to limit the required storage for preparing this dataset, the audio
is stored in the .flac format and is not converted to a float32 array. To convert, the audio
file to a float32 array, please make use of the `.map()` function as follows:
```python
import soundfile as sf
def map_to_array(batch):
speech_array, _ = sf.read(batch["file"])
batch["speech"] = speech_array
return batch
dataset = dataset.map(map_to_array, remove_columns=["file"])
```LCAR-Hallucination-Benchmark
LCAR Hallucination Benchmark
LCAR Hallucination Benchmark is a manually reviewed speech benchmark for
studying acoustic-grounding failures in LLM-based ASR. It contains two
500-utterance suites: controlled speech synthesized with IndexTTS2 and speech
derived from openly released corpora. The benchmark covers translation or
transliteration, spoken or text-prompt instruction execution, unsupported
repetition, and catastrophic deletion.
The benchmark is a targeted stress set. It is… See the full description on the dataset page: https://huggingface.co/datasets/aguangguang/LCAR-Hallucination-Benchmark.HALAS
Dataset Card for HALAS
Dataset Summary
HALAS (Hallucination Annotations for Large-scale ASR Systems) is a human-annotated dataset of hallucinations produced by modern automatic speech recognition (ASR) systems on real-world speech recordings. The dataset contains span-level hallucination annotations for ASR outputs generated from recordings in the Earnings22 corpus.
HALAS was introduced to address a key limitation in prior hallucination research: most existing… See the full description on the dataset page: https://huggingface.co/datasets/MatBar99/HALAS.knihi-be-janka_bryl_halia_all
AudioSet Pipeline Output
Мова / Language: Беларуская (Belarusian)
Аўдыё нарэзана з арыгінальнага запісу ў зыходнай частаце дыскрэтызацыі (native), мона, фрагменты да 30 секунд.
Частка калекцыі Belarusian Audiobooks (native).
Радкоў у датасеце
303
Працягласць
1 гадз
Частата дыскрэтызацыі
44100 Hz
Каналы
мона
Даўжыня фрагмента
да 30 с
Структура
Кожны радок змяшчае:
audio — аўдыёфрагмент (native SR, мона, ≤30 с)
text — транскрыпцыя (Gemini… See the full description on the dataset page: https://huggingface.co/datasets/fosters/knihi-be-janka_bryl_halia_all.el-mal-el-halal-podcast-subtitles
El Mal El Halal Podcast Subtitles
Dataset Summary
El Mal El Halal Podcast Subtitles is a collection of manual subtitles for 18 episodes of the El Mal El Halal podcast by Eng. Mohamed Aboulnaga, covering Arabic content. This dataset is designed for research on speech processing, translation, semantic search, and Arabic NLP.
Total episodes: 18 - untill the date of 03/08/2025
Total segments: 13 970
Total words: 166 505
Total duration: 20 h 50 m 56 s (75 057 s)
Average… See the full description on the dataset page: https://huggingface.co/datasets/hossam87/el-mal-el-halal-podcast-subtitles.knihi-be-janka_bryl_halia_output_original
AudioSet Pipeline Output — арыгінальнае аўдыё
Мова / Language: Беларуская (Belarusian)
Арыгінальнае аўдыё без апрацоўкі, захаванае ў зыходнай якасці.
Частка калекцыі Ministerskija —
корпус беларускіх аўдыёкніг.
Апрацаваная версія (сегменты ~15 с, выраўнаваная транскрыпцыя):
knihi-be-janka_bryl_halia_output
Структура
Кожны радок змяшчае:
audio — арыгінальны аўдыёзапіс
text — транскрыпцыя
chunk_uid — унікальны ідэнтыфікатар
Ліцэнзія / License
CC… See the full description on the dataset page: https://huggingface.co/datasets/fosters/knihi-be-janka_bryl_halia_output_original.