SYSTEM FOR LEARNING DICTIONARY FOR COOCCURRENCE RELATION OF WORDS

摘要

PURPOSE:To obtain a knowledge related to a cooccurrence relation by extracting common-occurrence relation data which are lacking from correction contents, registering them into a cooccurrence relation dictionary, extracting incorrect data in the cooccurrence relation dictionary and eliminating them, when a converting result is corrected by a user. CONSTITUTION:An input sentence is divided into words, the construction and the meaning of the sentence as the aggregation of the binomial relations of the word and the word are expressed, whether or not the combination of candidate translated words exist in the cooccurrence relation dictionary of a cooccurrence relation dictionary storage device 5 is checked for the binomial relations, when they exist, the combination is given with priority and selected and the translated word is determined. When the combination of the candidate translated words does not exist in the cooccurrence relation dictionary, the candidate translated word registered first into the vocabulary dictionary of a vocabulary dictionary storage device 4 is selected. When the converting result is inappropriate, the user executes a correction. Based on information extracted from the modification contents, data are accumulated into the cooccurrence relation dictionary and the incorrect data are eliminated from the cooccurrence relation dictionary.