To provide a content-retrieving device capable of retrieving contents making each user feel that recognition accuracy is high, taking the preference of the user into consideration.
The content-retrieving device has a sound feature quantity extraction means which inputs a voice waveform; a sound model accumulation means which outputs a sound probability P(x¾ω) for observing the sound feature quantity x for a word string ω of a recognition result candidate; a language model accumulation means which outputs a statistical/grammatical language probability Pn-gram(ω)/Pcfg(ω); a sound recognition decoder which outputs a recognition result word string ω^, based on x, P(x¾ω) and Pn-gram(ω)/Pcfg(ω); and a content retrieving means which retrieves contents from a content database by using ω as a retrieval key. The content retrieving result also has a preference degree probability calculation means which calculates the preference degree probability P
COPYRIGHT: (C)2010,JPO&INPIT
Masaki Naito
Tsune Kawai
JP2006085179A | ||||
JP2001282285A | ||||
JP2008204193A | ||||
JP2007164497A | ||||
JP2002523828A |