Scientists at the Institute of Smart Systems and Artificial Intelligence (ISSAI), located at the Nazarbayev University, have extended their work to an automated multilingual model that can recognize ten Turkic languages. Website of the university says.
Previously, scientists have developed speech recognition models for the Kazakh language. Now the system recognizes Azerbaijani, Bashkir, Chuvash, Kazakh, Kyrgyz, Sakha, Tatar, Turkish, Uighur and Uzbek languages.
«Multilingual models that were trained using general speech data showed more robust results than baseline monolingual models, with the best model achieving average symbol and word error rates of 56 and 54 percent, respectively,» the statement says.
Results of experiments showed that a reduction in character and word error rates was more likely when multilingual models were trained using data from related Turkic languages than when they were developed using data from unrelated, non-Turkic languages such as English and Russian.
To demonstrate the use of the multilingual automatic speech recognition model for Turkic languages, ISSAI has developed a demo program that recognizes ten Turkic languages, as well as Russian and English.
