Sanskrit Speech Recognition Corpus

84 hours of Sanskrit audio data for training automatic speech recognition models, uploaded by user 'addy88' to Hugging Face in December 2021. The dataset is categorized as containing 10K to 100K samples and includes text transcriptions.

Use Cases

• Train an ASR model to predict text transcriptions from Sanskrit audio waveforms.
• Fine-tune a pre-trained speech model on the 84 hours of Sanskrit audio for language adaptation.
• Analyze phonetic or acoustic features specific to Sanskrit speech from the audio samples.
• Benchmark ASR model performance on a dedicated Sanskrit speech corpus.

Strengths

• Contains 84 hours of dedicated Sanskrit speech data.
• Sized between 10,000 and 100,000 individual data samples.

Limitations

• Specific audio quality, speaker demographics, and recording conditions are unknown.
• The dataset has not been updated since 2021, which may limit its relevance for benchmarking against newer models.

Provenance

Source

Hugging Face user 'addy88'.

Freshness

Last updated in December 2021.