Author/Editor | Šef, Tomaž; Gams, Matjaž | |
Title | Obravnava homografov pri sintezi slovenskega govora | |
Translated title | Homograph disambiguation in Slovenian text-to-speech synthesis | |
Type | članek | |
Source | In: Zajc B, editor. Zbornik 9. elektrotehniške in računalniške konference ERK 2000. Zvezek B. Računalništvo in informatika, umetna inteligenca, robotika, razpoznavanje vzorcev, biomedicinska tehnika, močnostna elektrotehnika, didaktika, študentski članki; 2000 sep 21-23; Portorož. Ljubljana: Slovenska sekcija IEEE, | |
Publication year | 2000 | |
Volume | str. 169-72 | |
Language | slo | |
Abstract | Homograph disambiguation is a classification problem in which the output is a pronunciation label for an ambiguous target word. Slovenian language is a particulary rich source of homographs, due to lexical variations in stress patterns and the failure to distinguish e (narrow e ) from E (wide e) as well as o (narrow o) from O (wide o) in normal writing. In the morphological and pronunciation dictionary of 600.000 words (around 20.000 lemmas) we have collected close to 4.000 potential homographs that have been classified into several groups and subgroups.For each homograph, a large corpus containing labeled examples was constructed and analysed. The proposed algorithm was applied to several major types of ambiguity in which context can be used to choose the pronunciation of a word. | |
Descriptors | SPEECH ACOUSTICS PATTERN RECOGNITION LANGUAGE TESTS PHONETICS |