Meta STT Euro Set1: Metadata for European Speech Recognition Samples

European language speech recognition metadata sourced from CommonVoice and Multilingual LibriSpeech datasets. The dataset contains only metadata files in JSON or Parquet format, with audio files not included. It was created by WhissleAI and last updated on April 30, 2025.

Use Cases

• Benchmarking multilingual automatic speech recognition (ASR) systems based on metadata linking to CommonVoice and MLS audio.
• Analyzing speech data distribution and characteristics for European languages based on the described metadata.
• Preprocessing and filtering large-scale speech datasets based on metadata attributes before audio download.
• Training metadata-driven models for tasks like language identification or data quality assessment based on the linked samples.

Strengths

• Metadata is explicitly linked to two established, large-scale speech datasets: CommonVoice and Multilingual LibriSpeech.
• Covers multiple major European languages, including German, English, Spanish, French, Italian, and Portuguese.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

CommonVoice and Multilingual LibriSpeech datasets.

Collection Method

Aggregated metadata from the source datasets.

Freshness

Last updated 2025-04-30 05:13:49; freshness should be verified.

Geography

European languages.