Librispeech Audio Corpus for Speech Recognition

Librispeech39 is a subset of the LibriSpeech corpus, a widely used collection of read English speech derived from audiobooks. The dataset was uploaded to the Hugging Face platform by user 'arampacha' in May 2022. It is designed for training and evaluating automatic speech recognition systems.

Use Cases

• Train automatic speech recognition models to transcribe audio segments into text transcripts.
• Benchmark speech recognition model accuracy using standardized audio files and corresponding text.
• Develop audio feature extraction pipelines using the provided speech waveforms and phoneme alignments.

Strengths

• Derived from the well-established and widely cited LibriSpeech corpus, ensuring a known baseline.
• Uploaded to a major platform (Hugging Face) in 2022, facilitating access and integration with ML tools.

Limitations

• The specific size, number of hours, and exact composition of this '39' variant are unknown.
• Potential bias towards read, North American English speech from audiobook narrators.

Provenance

Source

LibriSpeech corpus, derived from public domain audiobooks from the LibriVox project.

Collection Method

Audio derived from read audiobooks, automatically segmented and aligned with text transcripts.

Freshness

Last updated on the platform in May 2022.

Geography

Primarily North American English speakers.