---
license: apache-2.0
datasets:
- mozilla-foundation/common_voice_17_0
language:
- lv
pipeline_tag: automatic-speech-recognition
---

# General-purpose Latvian ASR model

This is a fine-tuned [whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) model for Latvian, trained by [AiLab.lv](https://ailab.lv) using two general-purpose speech datasets: the Latvian part of [Common Voice 17.0](https://commonvoice.mozilla.org/lv/datasets), and a Latvian broadcast dataset [LATE-Media](https://korpuss.lv/id/LATE-mediji).

We also provide a 5-bit quantized version of the model in the GGML format.

## Training

Fine-tuning was done using the Hugging Face Transformers library with a modified [seq2seq script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition#sequence-to-sequence).

Training data:
| Dataset | Hours |
|:---|---:|
| Latvian CV 17.0 train set (the [V1 split](https://analyzer.cv-toolbox.web.tr/examine/lv/17.0)) | 167 |
| LATE-Media train set | 42 |
| Total | 209 |

## Evaluation

The model is evaluated on both a Latvian CV 17.0 test set (the [V1 split](https://analyzer.cv-toolbox.web.tr/examine/lv/17.0)) and a [LATE-Media test set](http://hdl.handle.net/20.500.12574/99).

| Dataset | WER | CER |
|:---|---:|---:|
| Latvian CV 17.0 V1 - formatted | 5.0 | 1.6 |
| Latvian CV 17.0 V1 - normalized | 3.4 | 1.0 |
| LATE-Media 1.0 - formatted | 20.8 | 8.2 |
| LATE-Media 1.0 - normalized | 14.1 | 5.9 |

## Citation

Please cite this paper if you use this model in your research:

```bibtex
@inproceedings{dargis-etal-2024-balsutalka-lv,
  author = {Dargis, Roberts and Znotins, Arturs and Auzina, Ilze and Saulite, Baiba and Reinsone, Sanita and Dejus, Raivis and Klavinska, Antra and Gruzitis, Normunds},
  title = {{BalsuTalka.lv - Boosting the Common Voice Corpus for Low-Resource Languages}},
  booktitle = {Proceedings of the Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING)},
  publisher = {ELRA and ICCL},
  year = {2024},
  pages = {2080--2085},
  url = {https://aclanthology.org/2024.lrec-main.187}
}
```

## Acknowledgements

This work was supported by the EU Recovery and Resilience Facility project [Language Technology Initiative](https://www.vti.lu.lv) (2.3.1.1.i.0/1/22/I/CFLA/002) in synergy with the State Research Programme project [LATE](https://www.digitalhumanities.lv/projekti/vpp-late/) (VPP-LETONIKA-2021/1-0006).
We are grateful to all the participants of the national initiative [BalsuTalka.lv](https://balsutalka.lv) for helping to make the Latvian Common Voice dataset much larger and more diverse.