System for Analysis and Synthesis of Tatar Speech

• normalization of the input phrase (deciphering abbreviations, dates, numerical information, etc.),
• phonemic transcription of the input sentence,
• semantic-syntactic analysis of the sentence,
• phonetic transcription of the synthesized phrase,
• prosodic formatting of the output phrase,
• audio rendering of the synthesized phrase.

Currently, databases of textual and speech information in the Tatar language are being accumulated and analyzed for development, machine learning technologies are being developed, and the integration of the Tatar language speech interface into modern PCs and mobile devices is underway. To create a universal speech recognition system, a database of voices from more than 400 speakers with a total duration of about 60 hours has been collected. The necessary programs and models have been created, and the first experimental version of the recognition system, which understands 200 thousand Tatar words, has been launched. The achieved results are comparable to world analogues and allow for "communication" with a computer using voice commands (speech translation, mobile assistants, message dictation, news reading).

• providing access to information for people with impaired vision;
• assisting communication for those with articulation disorders;
• use on the Internet: reading email, websites, knowledge bases;
• robotics and remote control;
• use in reference dialogue systems;
• application in mobile communication systems;
• automated language learning systems;
• use for voice-over of background texts;
• systems for audio-visual monitoring of instruments in conditions not accessible for direct observation, and others.