Librispeech Audio Transcription Metadata

Librispeech is a large-scale corpus of read English speech derived from audiobooks. The metadata for this dataset was uploaded by user leo19941227 to the Hugging Face platform in June 2022.

Use Cases

• Train acoustic models using the audio files and corresponding transcript text.
• Benchmark ASR system accuracy by comparing predicted transcript text to the ground truth labels.
• Analyze speaker characteristics or reading styles using metadata like speaker ID and chapter information.
• Perform text-to-speech alignment studies using the synchronized audio and transcription data.

Strengths

• Derived from the widely recognized and benchmarked Librispeech corpus.
• Contains approximately 1000 hours of speech across multiple quality tiers (clean, other).

Limitations

• Contains only read speech from audiobooks, lacking spontaneous conversational data.
• Primary source is English, limiting multilingual application.
• Metadata specifics like column names or exact row count are not provided in this upload.

Provenance

Source

Librispeech corpus, originally created by Vassil Panayotov and others.

Collection Method

Derived from public domain audiobooks from the LibriVox project.

Time Range

null

Freshness

Dataset uploaded in 2022; the original Librispeech corpus creation date is earlier.

Geography

null