| 摘要 |
<P>PROBLEM TO BE SOLVED: To register a registration candidate word of a natural language in a dictionary database without going through operator's assistance. <P>SOLUTION: A resource acquisition part 2 acquires natural language contents from an unfixed corpus according to an operation from an input part 1, a language analysis part 3 analyzes part-of-speech properties and modulation relation with other independent words by independent words of text data, and a language data measuring part 4 measures the frequencies of appearance of other independent words, having modulation relation with the independent words. A text data structure generating part 5 generates and stores the text data structure information showing the relation between the independent words and other independent words having the modulation relation with the independent words. An unregistered word attribute estimating part 7 temporarily impart a part-of-speech attribute to an unregistered word to be updated. An unregistered word evaluation part 8 receives text data structure information, regarding a candidate word to be registered and evaluates the text data structure information, based on prescribed standards, and a dictionary update part 9 removes the part-of-speech attribute and updates the candidate word to be registered as a registered word. <P>COPYRIGHT: (C)2005,JPO&NCIPI |
