| 摘要 |
PURPOSE:To improve the accuracy for extracting a sentence clause by constituting the system so that when an operator corrects a sentence clause delimiter, the sentence clauses before and behind the sentence clause delimiter before the correction, and in the subsequent conversion, the delimiter in this sentence clause is not selected, in the continuous sentence clause KANA (Japanese syllabary) to KANJI (Chinese character) conversion for using a keying interval time for a sentence clause delimiting processing. CONSTITUTION:With respect to a keying-inputted character-string, for instance, 'wa-ta- shi-wa-i...', the keying interval time Ti (i= 1-14) is stored. In conversion candidates 1, 2 of this character-string, for instance, the candidate 1 is assumed to be correct. Usually the candidate 2 is selected by a two-sentence clause longest coincidence method, and also, even by the delimiter candidate whose keying interval time is the longest, T3<T4 is set by a habit and erroneous keying of an operator, and in the case the candidate 2 is selected, the operator changes and corrects the candidate 2. In this case, sentence clause data of 'wa-ta-shi-wa' and 'i-shi-ya' are stored, referred to at the time of extracting the sentence clause from the next time, and that of the longest delimiting time is also excluded, and the sentence clause delimiter is selected from other candidate. |
