PART-OF-SPEECH GIVING DEVICE

摘要

PROBLEM TO BE SOLVED: To provide a device which can automatically and accurately give the parts of speech without using any part-of-speech dictionary. SOLUTION: A decision tree learning device 10 has a binary tree structure that is divided in dependence on every attribute value and based on the text data to which the parts of speech are already given according to plural attributes including the spelling features of each word, the features of words about how they are used in sentences and the hierarchical sorting using the mutual information contents of words. Then the device 10 generates a decision tree for giving the parts of speech and also generates a decision tree having the frequency provability by calculating the frequency probability of non-divided leaf nodes to plural parts of speech. A part-of-speech giving device 11 selects plural higher order ones of frequency provability given to the leaf nodes based on the input text data and by means of the decision tree having frequency probability. Then the device 10 gives those selected frequency provability to each word of the text data and decides a part-of-speech character that has the highest connection probability as a correct part-of-speech string among those word strings of the text data.