Search by:
Year of publication
Author name
Paper title
Lexical Stress-based Morphological Decomposition for Ukrainian Speech Recognition Tasks
Full text (PDF)
UDC: 004.934
Publication Language: Ukrainian
Stuc. intelekt. 2013; 18(3):233-241
Abstract: This paper presents an approach to the morphological level word segmentation based on lexical stress modeling, which is prospective for Ukrainian speech recognition systems. The formulated segmentation criterion is based on a training set of words with manually pointed stresses and a large text corpus. The described search algorithm finds one or more segmentations with the best likelihood. The developed toolkit is presented, experimental research is described and results are discussed.
Keywords: lexical stress, morphological decomposition, Ukrainian speech recognition
References:
- Black A. Issues in Building General Letter to Sound Rules . 3rd ESCA Workshop / Black A., Lenzo K., Pagel V.
- Creutz Mathias. Induction of a simple morphology for highly-inflecting languages / Creutz Mathias, Lagus Krista // In: Proc. 7th Meeting of the ACL SIGPHON. – 2004.
- Automatic lexical stress assignment of unknown words for highly inflected Slovenian language / [Gams Matjaž et al.] // In: Text, Speech and Dialogue. Springer Berlin Heidelberg, 2006.
- http://lcorp.ulif.org.ua/dictua/
- Bisani M. Joint-Sequence Models for Grapheme-to-Phoneme Conversion / M. Bisani, H. Ney // Speech Communication.
- Robeiko V. Bidirectional Text-To-Pronunciation Conversion with Word Stress Prediction for Ukrainian // V. Robeiko, M. Sazhok // In Proc. UkrObraz’2012.
- www.cybermova.com/speech/visual-hmm.htm
- Ukrainian Broadcast Speech Corpus Development / [Valeriy Pylypenko, Valentyna Robeiko, Mykola Sazhok, et al.] // In Proc. Specom’2011.