Kazakhstan Has Developed the Largest AI Model for Recognizing Turkic Languages

The largest automatic speech recognition (ASR) model for Turkic languages ​​was presented in Kazakhstan, The Caspian Post reports, citing Ministry of Artificial Intelligence and Digital Development Kazakhstan.

The Ministry of Artificial Intelligence and Digital Development, the development was presented by the startup Cybernet AI.

The project became the first full-scale AI model in Central Asia, originally created with the linguistic specifics of the Turkic group in mind, rather than adapted from English-language solutions.

The development was implemented with the support of Astana Hub and the Ministry of AI and Digital Development of the Republic of Kazakhstan using Microsoft's GPU infrastructure.

The model includes Kazakh, Turkish, Uzbek, Kyrgyz, Azerbaijani, and Tatar languages. The system also correctly recognizes mixed Turkic-Russian speech, which is widely used in everyday conversations, business communications, and when accessing government services.

Until now, Central Asian countries have primarily used foreign ASR models oriented towards the English language, which has reduced recognition accuracy due to the phonetic and accent features of Turkic languages.

Cybernet AI noted that the new model was created within and for the region, taking into account real-world spoken language.

The AI ​​solution is targeted at banks, telecom operators, contact centers, logistics companies, and government agencies. According to the developers, the model will reduce customer service costs and provide 24/7 voice support.

The system is resilient to noise, accents, and language switching within a single sentence, and demonstrates consistent accuracy across a variety of industries, from fintech to government services.