PURPOSE: To provide the dictionary data collector which speedily collect words in prescribed relation in a source word from the data of only an input source sentence, the vocabulary data of an existent input or the vocabulary data of the input source language sentence and the existent input according to an extremely simplified system completely without using a machine translation device or a dictionary for translation, etc.
CONSTITUTION: An input Japanese sentence is supplied to a pattern retrieval part 3 after being divided corresponding to the character type at a word division part 1. An extraction pattern storage part 2 stores plural extraction patterns, which is word patterns composed of the source word and the word in the prescribed relation with the source word, for extracting the word patterns, and the pattern retrieval part 3 extracts the extraction pattern coincident with the input word and outputs the pattern to a table editing part 4. The table editing part 4 classifies the words coincident with the extraction patterns for each pattern and further executes editing while detecting the appearance frequency of each word or the like.
MATSUDAIRA MASAKI
SAKAMOTO HITOSHI