datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
yt-kids-asr-bench
Child Speech ASR Benchmark
This dataset contains 16 kHz mono FLAC audio embedded in Parquet rows through the Hugging Face Audio feature. It is intended for manually reviewed ASR benchmarking access.
Configs
Config
Split
Rows
Duration
zh
test
18
02:24:00.076
en
test
55
03:43:24.454
v3
test
321
15:43:48.000
v3
zh
319
13:01:23.000
v4
en
55
05:29:50.418
v4
zh
104
05:09:14.000
total
872
45:31:39.948
Columns
audio: embedded… See the full description on the dataset page: https://huggingface.co/datasets/MagicLuke/yt-kids-asr-bench.CHILDES-Aligned
[!IMPORTANT]
How to access this dataset: the official public release is hosted by TalkBank at
https://talkbank.org/childes/access/Derived/CHILDES-Aligned.html (audio archives +
CSV/JSONL metadata, CC BY-NC-SA 4.0). Please obtain the dataset there.
This Hugging Face copy is retained gated, for internal use; access requests are
approved manually and general requests may be declined — use the TalkBank release instead.
CHILDES-Aligned: Curated Child-Speech Dataset (BEACON)
English… See the full description on the dataset page: https://huggingface.co/datasets/MagicLuke/CHILDES-Aligned.Redmond-Sentence-Recall
RSR segmented-latest
This dataset contains segmented child speech utterances built from the
RSR raw-latest CHAT/audio tree with talkbank-toolkit.
Initial upload target: MagicLuke/Redmond-Sentence-Recall (private).
Configs
Config
Split
Rows
Source manifest
sentence
train
14563
train.sentence.jsonl
sentence
test
2065
test.sentence.jsonl
all_chi
train
15102
train.all_chi.jsonl
all_chi
test
2167
test.all_chi.jsonl
Config Meaning… See the full description on the dataset page: https://huggingface.co/datasets/MagicLuke/Redmond-Sentence-Recall.