| 摘要 |
<p>PURPOSE:To eliminate the error in analysis due to an unknown word, and also, to make a back-track processing useless by using a kind of character information for a factor to determine the head of the unknown word. CONSTITUTION:An objective character string to be analyzed generation processing part 1 makes a character string, remaining after a part upto the end of the word, selected by a word selection processing part 3, is removed from an inputted writings, to be an objective character string to be analyzed. A word extraction processing part 2 prepares a candidate word list, by using a dictionary for a morpheme analysis 2a, a part of speech assorting list 2b, an inflection list 2c, a conjugation weight matrix 2d, and a paragraph end suitableness list 2e. When a candidate word list is not vacant, a word selection processing part 3 charges the word, just before the said list, in a word list, and when it is vacant, if two or more number of the characters except HIRAGANA (cursive Japanese syllabary) continues from the head of the said objective character string to be analyzed, the word selection processing part 3 starts to process the unknown word from its head character string. A determination processing part 4 generates a settled word list from the word list.</p> |
