| 摘要 |
PROBLEM TO BE SOLVED: To improve recognition performance by recognizing the input speeches delimited in words and uttered by referring to a word dictionary, forming candidate words with ranking, forming word candidate series from the candidate words for the respective input speeches, calculating the sum of scores based on the recognition ranking of the respective candidate words and displaying the word candidate series based thereon. SOLUTION: A speech recognition section 1 acoustically analyzes the voices uttered by a speaker together with a control section 5, references the word dictionary 6 and outputs the candidate words. A word system score computing section 2 searches a tree structure dictionary 3 from the combinations of the candidate words together with the control section 5, determines the respective candidate word series coinciding with the lists of the tree structure and determines the scores thereof. The control section 5 controls the individual processings necessary for speech recognition, score calculation, display of the recognition results, switching of the dictionaries, etc. A processing section 7 replaces the results of the speech recognition uttered in the word units with the desired character data array in the tree structure dictionary. A display section 4 displays the candidate word series of the recognition results or the selected word series. |
