ASTANA – The Til-Qazyna National Scientific-Practical Center has introduced the AI-powered Tilqazyna model for learning the Kazakh language. The large language model (LLM) can generate text, correct grammar errors, shorten sentences, create paraphrases, and analyze the context, reported the Ministry of Science and Higher Education on Jan 17.
It is available to users on the Hugging Face platform and will open up opportunities for creating IT products in the Kazakh language using artificial intelligence.
The center experts used natural language processing algorithms to create the model and analyzed large amounts of data, including 684,876 lexical units, 20,212-word combinations, 5,558 punctuation correction texts and 14,790 synonymous sequences. The total volume of the processed case was 35 gigabytes (GB).
Currently, the Tilqazyna supports levels A1 to B1, with plans to expand to B2 and C1 by 2026. The final product will be a mobile application with voice assistants that offer personalized learning programs.