ASR Language Model Training Corpus Version 2.1

Asr Book Lm V2.1 is a dataset for training automatic speech recognition language models, created by Jiejie. It was last updated on March 14, 2022.

Use Cases

• Training a language model for speech recognition using the provided text corpus.
• Fine-tuning an acoustic model with paired audio and text data for improved transcription accuracy.
• Benchmarking ASR system performance on a specific text corpus domain.
• Analyzing language model perplexity on speech recognition transcriptions.

Strengths

• Dataset is categorized as containing between 1k and 10k samples based on platform tags.
• Includes both text and audio modalities for multimodal training.

Limitations

• Specific sample count, file formats, and column structure are unknown.
• Data may be stale, with no updates since March 2022.

Provenance

Source

Jiejie (via Hugging Face)

Collection Method

null

Time Range

null

Freshness

Last updated in March 2022.

Geography

Region is tagged as 'us'.