Abror Shopulatov

I'm a Machine Learning Engineer at tilmoch.ai, where I mostly work on NLP applications for Turkic languages. At tilmoch.ai, I've worked on Tahrirchi Editor, tilmoch.ai Translator and Misollar.

We also contribute to the open-source Turkic NLP community with the biggest Uzbek text corpora (UzBooks and UzCrawl), the SOTA Uzbek encode models (tahrirchi-bert-base and tahrirchi-bert-small), Karakalpak and Southern Uzbek Machine Translation models and dataset.

At the same time, I am a BSc. in AI student at MBZUAI, where I am digging deeper into fundamentals of AI Engineering.

Email  /  CV  /  Scholar  /  Github  /  Huggingface

Research

I'm interested in NLP, Grammatical Error Correction and Machine Translation. Most of my research is on Uzbek and other Turkic languages..

With modern techniques in Machine Translation research, we have collected dataset and trained MT model for Southern Uzbek, a language with around 5 million speakers.

I created UzLiB, a comprehensive linguistic benchmark to evaluate LLMs on Uzbek. The results revealed that even top models fail to surpass 70% accuracy, highlighting a key gap in their capabilities.

We translated FLORES+ devset to Karakalpak language. Then, we collected 100,000 pairs involving Karakalpak to fine-tune machine translation model to improve upon existing baselines.

We developed a hybrid GEC model for Uzbek by integrating a rule-based morphological analyzer with a neural network, significantly improving its ability to handle complex agglutinative grammar.

We built and released UzBooks and UzCrawl by processing over 35,000 books and web data. At 36 GB, it is now the largest publicly available, high-quality text corpus for the Uzbek language.

Miscellanea

Feel free to steal this website's source code.