README.md

metadata

license: cc-by-nc-sa-4.0
language:
  - hsb
size_categories:
  - 1K<n<10K
task_categories:
  - automatic-speech-recognition
  - text-to-speech
tags:
  - audio
pretty_name: Dataset of Upper Sorbian speech recordings with transcriptions.

Upper Sorbian Speech Dataset: Korla & Cyril

This dataset contains approximately 8 hours of high-quality audio recordings of two native Upper Sorbian speakers, Korla and Cyril. It is designed primarily for speech synthesis and other speech-related tasks.

Dataset Summary

• Speakers: Korla and Cyril (native Upper Sorbian speakers)
• Total Duration: ~8 hours (the editing version includes less)
• Format: LJ Speech-style format (wav files and corresponding text)
• Alignment: Aligned using forced alignment, then manually checked and verified
• Text Correction: Efforts were made to correct transcripts of mispronounced words while preserving as much data as possible. However, no guarantees are made regarding grammatical or spelling accuracy.
• Recommended Use: Due to potential linguistic inconsistencies, this dataset is recommended only for speech synthesis (TTS).

Structure

The dataset contains two main folders:

• Cyril/ – Edited recordings of Cyril in LJ Speech format

• Korla/ – Edited recordings of Korla in LJ Speech format

• big .wav files: raw audio recordings

• big .txt files: the text that was spoken

License

This dataset is released under the Creative Commons Attribution-ShareAlike (CC-SA) license.

Citation

If you use this dataset in your research or projects, please credit the original authors appropriately and link to this repository.