Russian Crowdsourced Audio Transcription Benchmark

VoxDIY is a benchmark dataset for Russian automatic speech recognition, created via crowdsourcing. The dataset is monolingual (Russian) and was last updated in September 2024.

Use Cases

• Train an automatic speech recognition model on Russian crowdsourced audio transcriptions.
• Benchmark speech recognition system performance on monolingual Russian audio data.
• Analyze transcription patterns and potential label noise from crowdsourced annotations.

Strengths

• Dataset is specifically designed as a benchmark for Russian speech recognition.
• Annotations were created via crowdsourcing, potentially offering diverse transcription styles.
• The dataset was updated in September 2024, indicating recent maintenance.

Limitations

• The dataset size, row count, and column structure are unknown, limiting assessment of scale and features.
• Crowdsourced annotations may introduce label noise or inconsistency in transcriptions.
• Geographic or demographic bias may be present due to the crowdsourcing origin.

Provenance

Source

toloka

Collection Method

Crowdsourced audio transcription.

Freshness

Last updated 2024-09-10.

Geography

Region tag indicates 'us', but data language is Russian, suggesting potential geographic mismatch or focus.