Computer Science and Information Technologies, Computer Science and Information Technologies 2014

Font Size: 
Towards Automatic Speech Recognition for the Tatar Language
A. F. Khusainov, Dz. Sh. Suleymanov

Last modified: 2021-02-27

Abstract


In this paper we describe an approach to create automatic speech recognition systems for the Tatar language. We developerd speech analysis platform to work with under-resourced languages and used this tool to create baseline speech recognition system. Additionally, some changes have been made to this language-independent system to take into account specific Tataer morphological structure. The resulting adapted system showed 75% accuracy on testing audio records.

Keywords


Tatar language; under-resourced languages; recognition system

References


1. Lewis, M. Paul, Gary F. Simons, Charles D. Fennig (eds). "Ethnologue Languages of the World", Dallas, Texac: SIL International, 2013.

2. Khusainov A.F. "Automatic phoneme recognition system for the Tatar language". IN: The 1st Intertational Conference "TurkLang". Astana, 2013, pp 211-217.

3. Young S. Kershaw D., Odell J., Ollason D., Vatchev V., Woodland Ph. The HTK Book [Electronic resource]. URL: http://nesl.ee.ucla.edu/project/ibadge/ASR/htk/htkbook.pdf.

4. Kurimo M. Puurula A., Arisoy E., Alumae T., Saraclar M., "Unlimited vocabulary speech recogtion for agglutinative languages". In: HLT-NAACL, NY, USA. 2006, pp 487-494

Full Text: PDF