Common Voice ASR Clean: Filtered Speech Recognition Samples

A filtered version of the Common Voice dataset for automatic speech recognition (ASR). Samples with fewer than three words, repetitive tokens, or chat token leaks have been removed. The dataset was created by OpenSpeechHub and was last updated on March 31, 2026.

Use Cases

• Train automatic speech recognition models based on filtered audio-text pairs.
• Benchmark ASR model performance on a dataset with removed short and repetitive samples.
• Fine-tune language models for speech tasks using cleaned transcriptions.

Strengths

• Applies specific filtering criteria: removes samples with fewer than three words, repetitive tokens, or chat token leaks.
• Last updated on March 31, 2026, suggesting recent maintenance.

Limitations

• Row count, file formats, and column-level documentation are unknown, limiting suitability assessment.
• License information is unavailable, which is critical for determining permissible use.

Provenance

Source

OpenSpeechHub, likely derived from the Mozilla Common Voice project.

Collection Method

Filtered from a source ASR dataset by removing samples meeting specific criteria.

Time Range

null

Freshness

Last updated 2026-03-31 18:13:56

Geography

null