| 摘要 |
PROBLEM TO BE SOLVED: To generate a pattern set to be used by automatically analyzing learning data, when a name is specified by pattern matching. SOLUTION: A data conversion part 11 inputs learning data from a learning data storage part 12 and analyzes the parts of speech of individual morphemes to convert the data into morpheme data with name tags indicating the kinds of parts of speech by morpheme in the name tags. A pattern candidate generation part 13 extracts combinations of parts of speech of a morpheme array from the learning data and registers the combinations of parts of speech as pattern candidates in a pattern candidate storage port 14. Further, a pattern candidate evaluation part 15 extracts a morpheme array which matches a pattern candidate from the learning data, counts the frequency of the morpheme array appearance in the learning data and the frequency as to its name for finding an evaluated value, and registers the pattern candidate as a pattern in a pattern storage part 16 based on the evaluated value to generate a pattern set. |
