Saar-Voice: Multi-Speaker Speech Corpus for Saarländisch Dialect

9 speakers contributed roughly 6 hours of audio recordings in the Saarländisch dialect of German. The corpus contains 4,871 recorded sentences and an additional 3,905 unrecorded sentences, sampled at 22,050 Hz. It was created by UdS-LSV and last updated in June 2026.

Use Cases

• Train automatic speech recognition models for the Rhine Franconian dialect based on the 6 hours of audio.
• Develop text-to-speech systems for Saarländisch based on the multi-speaker recordings.
• Conduct linguistic analysis of dialectal variation using the corpus of recorded and unrecorded sentences.
• Benchmark speaker identification or diarization algorithms using the 9 distinct speaker recordings.

Strengths

• Contains 4,871 recorded sentences, providing a substantial base of audio data.
• Includes 9 distinct speakers (P01–P09), enabling multi-speaker analysis.
• Focuses on the Saarländisch dialect, a specific and likely underrepresented linguistic variety.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The split between recorded and unrecorded sentences suggests incomplete coverage of the textual corpus.

Provenance

Source

UdS-LSV

Freshness

Last updated 2026-06-16 17:21:19; freshness should be verified.

Geography

Saarbrücken and the surrounding region in Germany.