Abstract
In this paper, a keyword spotting based dialogue system is described. It is critical to understand user's requests accurately in a dialogue system. But the performance of large vocabulary continuous speech recognition (LVCSR) system is far from perfect, especially for spontaneous speech. In this work, an improved keyword spotting scheme is adopted instead. A fuzzy search algorithm is proposed to extract keyword hypotheses from syllable confusion networks (CN). CNs are linear and naturally suitable for indexing. To accelerate search process, CNs are pruned to feasible sizes. Furthermore, we enhance the discriminability of confidence measure by applying entropy information to the posterior probability of word hypotheses. On mandarin conversational telephone speech (CTS), the proposed algorithms obtained a 4.7% relative equal error rate (EER) reduction.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
