摘要 |
PROBLEM TO BE SOLVED: To provide a morpheme analyzer capable of automatically giving a part of speech by judging words and non-words with respect to an inputted sentence, which is not separately written, so as to divide into each word. SOLUTION: A deciding tree learning device 10 is provided with the tree structure of respective halved tree systems to generate a part-of-speech deciding tree for giving the part of speech and a word deciding tree for dividing a word. A word dividing and a part-of-speech giving device 11 selects plural high-ordered frequency probability in the frequency probability of a word category give to the leaf node of the word deciding tree by using the word deciding tree, based on inputted text data to give them to each word candidate, selects plural high-ordered frequency probability in the frequency propability of a part-of- speech category given to the leaf node of the part-of-speech deciding tree by using the part of speech deciding tree based on text data to give them to each word candidate and decides the column of the combination of words and the parts of speech divided into words, which is provided with maximum combining probability at the word column of text data, as the column of right answers. |