nur-dev/ait-asr

AitASR

AitASR is a fine-tuned version of OpenAI's Whisper Small model for Automatic Speech Recognition (ASR) in the Kazakh language. It was trained on the farabi-lab/kazakh-stt dataset to improve transcription quality for Kazakh audio.

---

🔧 Intended Use

The model is designed for ASR tasks involving Kazakh-language audio.
It is suitable for:

• Transcription of Kazakh speech
• Voice command recognition
• Speech-driven applications in Kazakh

---

⚠️ Limitations

• May perform poorly on:
  • Low-quality or noisy audio
  • Audio from domains significantly different from the training data
• Not optimized for real-time use without further engineering

5. Citation

If you use this model, please cite it as follows:

@article{kadyrbek2023ksd,
  author = {Kadyrbek, N.; Mansurova, M.; Shomanov, A.; Makharova, G.},
  title = {The Development of a Kazakh Speech Recognition Model Using a Convolutional Neural Network with Fixed Character Level Filters},
  journal = {Big Data and Cognitive Computing},
  year = {2023},
  volume = {7},
  number = {3},
  pages = {132},
  doi = {https://doi.org/10.3390/bdcc7030132}
}```

---
Commercial Use
For commercial use, please contact the author directly to discuss licensing terms and permissions.

<!-- measured-benchmark:start -->
## Evaluation (independently measured)

Held-out public test sets, measured directly — **not** self-reported (seed 42,
uniform multilingual-Whisper normalization). FLEURS test = 500 utterances/language;
ISSAI KSC2 test = 1000 utterances (in-domain Kazakh, spanning
crowd/parliament/podcasts/radio/talkshow).

| Test set | Lang | WER (%) | CER (%) |
|---|:--|--:|--:|
| FLEURS `kk_kz` | kk | 17.10 | 7.60 |
| FLEURS `ru_ru` | ru | 76.16 | 39.61 |
| FLEURS `en_us` | en | 39.42 | 30.82 |
| ISSAI KSC2 | kk | 36.05 | 15.35 |

Macro WER (kk/ru/en): **44.23%** (unweighted mean; penalises models that do not
cover all three languages).

> **Note.** The card reports no numbers. Measured: usable for Kazakh (FLEURS 17.1); Russian and English are weak (76 / 39). Same weights as ait-asr-kazakh.
<!-- measured-benchmark:end -->

<!-- license-note:start -->
## License & commercial use

**Non-commercial use only** (CC BY-NC 4.0). For commercial licensing or other
inquiries, please reach out to the author, **Nurgali Kadyrbek**, on LinkedIn:
https://www.linkedin.com/in/nurgali-kadyrbek-504260231/
<!-- license-note:end -->