| 摘要 |
PURPOSE:To provide a compound word dictionary registering device capable of automatically discriminating whether or not a word string is to be registered as one compound word, appropriately suppressing the size of a dictionary and improving the accuracy of a natural language analysis processing. CONSTITUTION:The compound word inputted from a keyboard 10 and a file 11 through an input part 9 is read and separated into words by a word separation part 4, a dictionary 2 is referred to for the separated character string and the frequency of the appearance is found. An evaluation value calculation part 6 calculates an evaluation value based on the frequency of the appearance. A registration deciding part 7 decides the compound word to be registered based on the calculated evaluation value. Since the compound word to be registered is decided based on the evaluation value calculated based on the frequency of the appearance, the compound word to be registered can be automatically discriminated. Also, by being applied to a machine translation processing or a KANA (Japanese syllabary)/KANJI (Chinese character) conversion processing, correct translated words and correct converted results can be obtained. |
