Multi-Speaker Sinhala Audio with Manual Quality Checks

Featuring multi-speaker, high-quality transcribed audio data for the Sinhala language, consisting of wave files and a TSV file. The data was manually quality checked and was collected by Google in Sri Lanka and contributed by the Path to Nirvana organization.

Use Cases

• Train a text-to-speech model using the high-quality audio wave files and their corresponding transcriptions.
• Develop a multi-speaker voice synthesis system leveraging the speaker information embedded in the FileID structure.
• Build a Sinhala speech recognition model by training on the paired audio and transcription data.
• Analyze speaker-specific phonetic characteristics using the UserID and audio file associations.

Strengths

• Data has undergone manual quality checking.
• Contains multi-speaker audio, providing speaker diversity.
• Includes high-quality transcribed audio specifically for the Sinhala language.

Limitations

• The dataset description acknowledges there may still be errors despite manual checks.
• Key metadata such as row count, column count, file size, and license are unknown.
• The last update was in October 2021, which may limit relevance for current applications.

Provenance

Source

Collected by Google in Sri Lanka and contributed by Path to Nirvana organization.

Collection Method

Collection and contribution as described; specific gathering method unknown.

Freshness

Last updated on 2021-10-04.

Geography

Sri Lanka (Sinhala language).