Datasetpublic

akiva-skolnik/hebrew-impairment-speech-v1

Hebrew Atypical Speech Dataset (Down Syndrome) Dataset name: hebrew-impairment-speech-v1 Speaker: A single Hebrew speaker with Down syndrome. Purpose: To advance research in personalized ASR for atypical speech, especially Hebrew. Dataset Summary This dataset contains 2307 Hebrew audio clips spoken by one individual with Down syndrome, accompanied by transcriptions. Speech includes dysarthric, stuttered, and non-standard pronunciation and grammar. The dataset aims… See the full description on the dataset page: https://huggingface.co/datasets/akiva-skolnik/hebrew-impairment-speech-v1.

sourceHugging Facecc-by-nc-4.0updated 9mo agoView on Hugging Face

1likes98downloads

Dataset Card

Hebrew Atypical Speech Dataset (Down Syndrome)

Dataset name: hebrew-impairment-speech-v1 Speaker: A single Hebrew speaker with Down syndrome. Purpose: To advance research in personalized ASR for atypical speech, especially Hebrew.

Dataset Summary

This dataset contains 2307 Hebrew audio clips spoken by one individual with Down syndrome, accompanied by transcriptions. Speech includes dysarthric, stuttered, and non-standard pronunciation and grammar.

The dataset aims to support research into:

—Personalized ASR
—Hebrew impaired-speech recognition
—Assistive communication systems

Approved for release by the speaker’s family for non-commercial use.

Data Structure

audio/folder/*.wav
dataset.jsonl  // {"file": "audio/0001.wav", "text": "…"}

Statistics

Metric	Value
Clips	2307
Total duration	4655.7 sec (~1.29 hours)
Min	0.286 sec
Max	37.989 sec
Average	2.02 sec
Median	1.03 sec

Example: Whisper Fine-Tuning Notebook

See notebook.ipynb