Librispeech Audio Metadata

Librispeech Metadata provides descriptive information for the LibriSpeech audio corpus, a widely used benchmark in speech recognition. The metadata was uploaded by s3prl to the Hugging Face platform in June 2022. It serves as a companion to the primary audio dataset.

Use Cases

• Analyze speaker demographics or recording characteristics for bias assessment in ASR systems.
• Filter audio samples by metadata attributes like speaker ID or chapter for targeted model training.
• Correlate audio file performance metrics with metadata fields to identify challenging data subsets.

Strengths

• Metadata is associated with the LibriSpeech corpus, which contains approximately 1000 hours of speech.
• Uploaded by a known contributor (s3prl) on a major ML data platform.

Limitations

• Specific column names, row counts, and the exact scope of the metadata are unknown.
• The metadata's recency is unclear as the underlying LibriSpeech corpus was released earlier.

Provenance

Source

Derived from the LibriSpeech audio corpus.

Collection Method

Likely extracted or compiled from the original LibriSpeech distribution.

Time Range

null

Freshness

Last platform update was in 2022.

Geography

null