| 摘要 |
<P>PROBLEM TO BE SOLVED: To automatically extract hierarchical structure information of vocabulary from a corpus of words, documents, etc., in which morpheme information is connected. <P>SOLUTION: A two-word hierarchical relation deriving means 11 predicts an inclusion relation with other morpheme having a connection relation for each morpheme from an aggregate of documents, etc., in which a morpheme of a corpus 2 is connected, calculates an inclusion rate by using a statistical index, derives a hierarchical relation between two words from the inclusion rate, and stores them in information base 13. A hierarchical structure construction means 15 extracts a hierarchical relation between tow words having the same morpheme in the base 13 as a superordinate word, sets the hierarchical relation between the two words having the highest inclusion rate of the superordinate word as an initial hierarchy, selects the hierarchical relation between two words in which a subordinate word of the set two-word hierarchical relation is the superordinate word and the inclusion rate is the highest one and connects the selected hierarchical relation as a subordinate hierarchy of the initial hierarchy, and outputs vocabulary hierarchical structure information 3 with the extracted two-word hierarchical relation connected thereto. <P>COPYRIGHT: (C)2008,JPO&INPIT |
