BTU Completes Another Milestone in Developing Georgian Language Resources for Artificial Intelligence – 150 Million Tokens Prepared

Business and Technology University (BTU) has completed a significant stage in the development of Georgian language training resources for artificial intelligence.

By creating and openly publishing Georgian-language datasets, grammatical resources, and training examples, the university has contributed to expanding the linguistic foundation necessary for the use of Georgian in AI systems.

This work is essential to enabling people to interact more effectively with artificial intelligence in Georgian – to learn, work, search for information, and create content. The availability of high-quality digital resources for the Georgian language will determine how widely these technologies can be used in education, business, research, and everyday life.

The core linguistic material in BTU’s new dataset contains approximately 150 million tokens. The new version builds on previously published resources and significantly expands the materials available for training and evaluating AI systems in Georgian.

The dataset contains millions of linguistic entries and examples developed according to Georgian grammatical rules. These include various grammatical structures, context-specific forms, and sentence construction patterns.

During earlier stages of the project, BTU published two Georgian-language datasets on GitHub, with a combined volume of approximately 27 million tokens.

The university also made a comprehensive Georgian grammar resource for AI publicly available. The new version continues this work, providing developers and researchers with broader opportunities to improve the understanding, processing, and generation of Georgian text.

The project’s partner is Palitra Group, which provided BTU with a large collection of Georgian-language texts. These materials, together with the university’s own resources, form the foundation for processing linguistic data and developing new resources.

The addition of this new resource creates opportunities to continue improving the use of Georgian in AI systems by leveraging a more extensive collection of training materials. One of BTU’s key objectives is to see further improvements in Georgian-language capabilities across various AI systems by the end of the year.

The dataset is available on BTU’s GitHub page.

By completing this stage, the university is providing technology companies working with the Georgian language with additional resources that can be used to develop new Georgian-language AI tools and improve existing systems.