The worlds first large language model for the Tibetan language was unveiled in China
In Lhasa, the administrative center of the Tibet Autonomous Region, the world’s first large-scale language model in the Tibetan language—DeepZang—was unveiled. It was developed to provide real-time translation and answer questions in Tibetan, marking an important step in the development of technologies for China’s minority languages.
Source: Globaltimes
The model was developed by CHOKNOR Information Technology, led by Tenzin Norbu. DeepZang supports over 80 languages, including Mandarin, English, Mongolian, and Uyghur. This model combines speech recognition, translation, and text generation capabilities, enabling a wide range of applications.
Immediately after its official launch, the service became very popular—it was accessed an average of 4,000 times per hour. To train the model, a corpus of 70 million pairs of Tibetan and Chinese texts was used, and the country’s first Tibetan language database was created, covering three major dialects.
CHOKNOR Information Technology specializes in developing innovative technologies in the field of artificial intelligence, particularly in the area of language models and translation for various national languages of China. Its goal is to promote the preservation and development of minority languages through modern technologies.
This project has significant cultural and technological implications, as it not only helps preserve the Tibetan language but also integrates it into the modern digital world. In the long term, the development of such models will help ensure access to information and communication for numerous linguistic communities.
This is just the beginning of a major effort to harness China’s linguistic resources, and we can expect further improvements in technologies that support minority languages, expanding their role in the global information space.
