Razpoznavanje govora v domeni dnevno-informativnih oddaj / Speech Recognition in the Broadcast News Domain
Ključne besede: tekoči govor; razpoznavanje govora; pregibni jezik.
Abstract: In this paper the structure of a speech recognition system is presented. We focus on Slovenian speech recognition. UMB BN system is described. Currently, it is the most complex large vocabulary continuous speech recognition system for Slovenian language. It is designed for transcribing TV-news shows. In the paper language resources are also described being crucial in developing such systems.
Key words: speech recognition; highly inflective language; language resources