This latest development is capable of accurately recognizing several languages, including Kazakh, Uzbek, Kyrgyz, Turkish, Tatar, and Azerbaijani, as well as combinations of Turkic-Russian speech. This will significantly improve the quality of voice services for both businesses and government institutions.
Previously, Central Asian countries used ASR models focused on the English language, which negatively affected recognition accuracy due to the unique features of Turkic phonetics and pronunciation. The new model, created specifically for live communication, is resistant to background noise and can adapt to various accents and language switches within a single utterance.
The project was implemented based on Microsoft’s GPU infrastructure with the support of Astana Hub and the Ministry of Digital Development, Innovations, and Aerospace Industry of the Republic of Kazakhstan.
