摘要 |
PURPOSE: To provide a natural language syntax analyzing device which can take an analysis at a higher speed with higher precision by greatly decreasing combinations of words (many meanings and multiple-part-of-speech word) whose dependency relation is to be decided while maintaining simple analytic algorithm. CONSTITUTION: A dependency relation judging part 20 judging dependency relations between two adjacent words or subtrees on the basis of a dependency relation judging rule set 21 sequentially from the sentence head side of word information of a morpheme array having a point of interest controlled by an interest point control part 10, and a many meanings and multiple-part-of-speech selective decision part 30 judges the possibility of the presence of another solution and its priority relation by referring to adjacent subtrees or words and many meaning and multiple-part-of-speech information in addition to a subtree or word of interest on the basis of a many meaning and multiple-part-of-speech selection rule set 31, and a many meaning and multiple-part-of-speech exchange part 40 performs an exchanging process according to the judging result, a subtree generation part 50 generates a subtree, and an analytic tree output part 60 outputs an analytic tree where the words from the sentence head to the tail are put together in one. |