摘要 |
PROBLEM TO BE SOLVED: To enable extrapolation response with user utterance before the end of the same. SOLUTION: Response decision processing is executed by subjecting the user utterance 'AIKOU...ISHIDA' to word recognition like heretofore, forming a word lattice, rebuilding the word lattice from the word token thereof. A phoneme lattice is formed by phoneme recognition which is the source for the word recognition and the word token 606 including phoneme position information is formed in accordance with the phoneme lattice and is sent to a response decision section. 'AIKOU ISHIDA' is recognized by the response decision processing 6706, 6710, 6713 for the tokens 6604, 6608, 6610 for 'AIKO'. 'ISHIDA' 692 is then extrapolatively response uttered. |