|Article title||ON RECOGNITION OF THE SUPER DIMENSIONED VOCABULARIES OF RUSSIAN WORD FORMS USING QUASI-WORD STEMS|
|Authors||V.Yu. Shelepov, A.V. Nicenko|
|Section||SECTION II. CONTROL SYSTEMS, SIMULATION AND ALGORITHMS|
|Month, Year||04, 2016 @en|
|Abstract||The paper discusses the method of word recognition of the super dimensioned vocabularies of Russian word forms using the authors’ automatic segmentation of speech signal system. The method can be applied both for DTW-recognition and recognition by means of the Hidden Markov models. But we mean method of diphone DTW- recognition (developing by authors) when practical recognition of speech units is described. Hence we adduce the conception of formalized diphone. Then the notions of quasi-word stem and quasi-inflection are defined. The general algorithm of quasi-word stems construction for the given list of word forms is suggested. The algorithm of Russian participles recognition using quasi-word stems is expounded. We suggest the recognizing algorithm of the words initial sounds (or sufficiently narrow classes of they belong to) in order to accelerate the recognition of the large vocabulary of quasi-word stems. In conclusion it should be noted that there is an objective problem of robust recognition of short words. Therefore quasiword stems have good recognition when they are sufficiently long. In general case the vulnerable point of quasi-word stems using is the recognition of shorter speech segments in comparison with the primary words (the step to overcome this difficulty lies in above-mentioned procedure of the first sound classification). Nevertheless using of quasi-word stems seems reasonable in recognition of the super dimensioned vocabularies of Russian word forms.|
|Keywords||Segmentation, quasi-word stem, parts of the first sound, recognition classes, transitions depending on intermediate recognition results|
|References||1. Buribaeva A.K., Dorokhina G.V., Nitsenko A.V., Shelepov V.Yu. Segmentatsiya i difonnoe raspoznavanie rechevykh signalov [Divonne segmentation and recognition of speech signals], Trudy SPIIRAN [Proceedings of SPIIRAS], 2013, Issue 31, pp. 20-42.
2. Mporas I., Ganchev T. and Fakotakis N. Speech segmentation using regression fusion of boundary predictions, Computer Speech and Language, 2010, Vol. 24, No. 2, pp. 273-288.
3. Gуmez J.A., Calvo M. Improvements on Automatic Speech Segmentation at the Phonetic Level, Progress in Pattern Recognition, Image Analysis, Computer Vision, and Applications, 2011, Vol. 7042, pp. 557-564.
4. Natarajan V.A., Jothilakshmi S. Segmentation of Continuous Speech into Consonant and Vowel Units using Formant Frequencies, International Journal of Computer Applications, 2012, Vol. 56, No. 15, pp. 24-27.
5. Yuan J., Ryant N., Liberman M. [and all]. Automatic phonetic segmentation using boundary models, Proceedings of Interspeech, 2013, pp. 2306-2310.
6. Patc Z., Mizera P., Pollak P. Phonetic Segmentation Using KALDI and Reduced Pronunciation Detection in Causal Czech Speech, Text, Speech, and Dialogue, 2015, Vol. 9302, pp. 433-441.
7. Zhijian Ou, Xiao Ji. A study of large vocabulary speech recognition decoding using finite-state graphs, Chinese Spoken Language Processing (ISCSLP), 7th International Symposium, 2010, pp. 123-128.
8. Susman D., Kopru S., Yazici A. Turkish Large Vocabulary Continuous Speech Recognition by using limited audio corpus, Signal Processing and Communications Applications Conference, 2012, pp. 1-4.
9. Saon G. Jen-Tzung Chien. Large-Vocabulary Continuous Speech Recognition Systems: A Look at Some Recent Advances, Signal Processing Magazine, 2012, Vol. 29, No 6, pp. 18-33.
10. Byrne W., Hajič J., Ircing P. [and all]. Morpheme Based Language Models for Speech Recognition of Czech, Text, Speech and Dialogue, 2002, Vol. 1902, pp. 211-216.
11. Saraswathi S., Geetha T. Morpheme based language model for Tamil speech recognition system, The International Arab Journal of Information Technology, 2007, Vol. 4, No. 3, pp. 214-219.
12. Rotovnik T., Maučec M.S., Kačič Z. Large vocabulary continuous speech recognition of an inflected language using stems and endings, Speech Communication, 2007, Vol. 49, No. 6, pp. 437-452.
13. Hong Kai Sze, Tan Tien Ping, Tang Enya Kong, Cheah Yu-N. Linguistic stem concatenation for malay large vocabulary continuous speech recognition, Research and Development (SCOReD), 2010 IEEE Student Conference on, 2010, pp. 144-148.
14. Karpov, I. Kipytkova, A. Ronzhin. Very large vocabulary ASR for spoken Russian with syntactic and morphemic analysis, Proceedings of INTERSPEECH’ 2011, Florence, 2011, pp. 3161-3164, pp. 291-296.
15. Stas J., Hladek D., Juhar J., Zlacky D. Analysis of morph-based language modeling and speech recognition in Slovak, Information and communication technologies and services, Vol. 10, No. 4, pp. 2012, special issue.
16. Shelepov V.Yu., Nitsenko A.V. K probleme raspoznavaniya slitnoy rechi [To the problem of continuous speech recognition], Iskusstvennyy intellect [Artificial intelligence], 2012, No. 4, pp. 272-281.
17. Shelepov V.Yu., Nitsenko A.V. O nekotorykh voprosakh, svyazannykh s difonnym raspoznavaniem i raspoznavaniem slitnoy rechi [On some issues related givonim, recognized the existence and the continuous speech recognition], Iskusstvennyy intellect [Artificial intelligence], 2013, No. 3, pp. 209-216.
18. Nicenko A.V. A «by part» method of Russian word speech recognition, Eurasian Journal of Mathematical and Computer Applications, 2014, Vol. 1, Issue 2, pp. 102-109.
19. Zaliznyak A.A. Grammaticheskiy slovar' russkogo yazyka [Grammatical dictionary of the Russian language]. Moscow: Russkiy yazyk, 1977, 879 p.
20. Dorokhina G.V. Pavlyukova A.P. Modul' morfologicheskogo analiza slov russkogo yazyka [The module of morphological analysis of words of the Russian language], Iskusstvennyy intellect [Artificial intelligence], 2004, No. 3, pp. 636-642.
21. Patent Ukraїni № 78806 «Pristrіy dlya zberezhennya і poshuku ryadkovikh velichin ta sposіb zberezhennya і poshuku ryadkovikh velichin» Vlasnik: Іnstitut problem shtuchnogo іntelektu [The patent of Ukraine No. 78806 "Device for storage and retrieval of string values and a way of saving and finding a string value" Owner: Institute of problems of artificial intelligence], Vinakhіdnik Dorokhіna G.V., Promislova vlasnіst' [Industrial property]. Byul. No. 5.
22. Buribayeva A.K., Sharipbay A.A. Kazakh Vowel Recognition at the Beginning of Words, Mediterranean Journal of Social Sciences, MCSER Publishing, Rome-Italy, 2015, Vol. 6, No 2, S4, pp. 121-127.