Disfluency and Out-Of-Vocabulary Word Processing in Arabic Speech Understanding
Younés Bahou, Lamia Hadrich Belguith, Abdelmajid Ben Hamadou · 2008
The disfluencies inherent in spontaneous speaking and out-of-vocabulary words omnipresent in any transcribed oral utterance by speech recognition, are a real challenge for speech understanding systems. Thus, we propose in this paper, a method for processing disfluencies and out-of-vocabulary words in the context of automatic Arabic speech understanding. Our method based on a robust and partial analysis of Arabic oral utterances (conceptual segments analysis) is effective for the treatment of such phenomena. This method has been tested through the understanding module of SARF system, an interactive vocal server for Tunisian railway information.