The recognition vessel weight study device and the speech input system, and system

摘要

The speech input system (110), the optimum recognition result is selected gathering of the speech recognition vessel (s1-sM) from the recognition result which is output due to majority. The study device (100) adding the weight value which was studied in regard to gathering of the speech recognition vessel, it does this majority. The study device has the expedient which seeks the recognition result of sound for study sound for study (101) the expedient which selects the speech recognition vessel which corresponds to quality (103) with, with the speech recognition vessel which is selected (104) with, integrating the recognition result, the expedient which forms word line network (105) with, the expedient which calculates the weight value regarding gathering of the speech recognition vessel with study (106) with. The study device, when calculating weight value, selects word due to the majority which adds weight value candidacy from each arc gathering of word line network, outputs the weight value candidacy where recognition error ratio in the word line which consists of the word which is selected becomes smallest as the study result.