CoolFace
Datasetpublic

google/svq

Simple Voice Questions Simple Voice Questions (SVQ) is a set of short audio questions recorded in 26 locales across 17 languages under multiple audio conditions. It serves as a core evaluation componenet for Massive Sound Embedding Benchmark (MSEB). Technical Specifications Feature Details Locales 26 Languages 17 Total Speakers ~700 (Capped at 250 recordings per speaker) Audio Conditions Clean, Background Speech, Media, Traffic Noise Gender… See the full description on the dataset page: https://huggingface.co/datasets/google/svq.

sourceHugging Facecc-by-4.0updated 3mo agoView on Hugging Face
58likes68kdownloads
filedocument_retrieval_cross_lang.parquet10.7 MBdownload
filedocument_retrieval_in_lang.parquet13.5 MBdownload
filepassage_retrieval_cross_lang.parquet7.2 MBdownload
filepassage_retrieval_in_lang.parquet9.6 MBdownload
filequery_reranking.parquet648.8 MBdownload
filesalient_term.parquet7.5 MBdownload
filespan_reasoning_cross_lang.parquet24.7 MBdownload
filespan_reasoning_in_lang.parquet57.8 MBdownload
filespan_retrieval_cross_lang.parquet11.0 MBdownload
filespan_retrieval_in_lang.parquet13.9 MBdownload
filespeech_transcription.parquet10.1 MBdownload

google/svq · main · files are served by the source, never re-hosted here