To provide an ambiguity solving device for natural language, capable of easily and reliably solving ambiguity involved in natural language.
The ambiguity solving device 50 comprises a retrieval part 100 which collects, for each of combinations (W, Rk) of a word W in an input sentence with semantic candidates Rk in a semantic candidate set corresponding to the word, a group of documents with co-occurrence of these words from the Internet 52; a decision tree formation part 80 which automatically forms, when the word W and a characteristic quantity vector showing a context in the document of the word are given, a decision tree 82 for selecting a semantic candidate most suitable as the meaning in the context of the word W from the set of semantic candidates by use of the collected document group as learning data; and a classification characteristic vector formation part 84 and a classification execution part 86 which use the characteristic vector obtained from the context of the word W, in the input sentence, to select an optimum one as the meaning of the word W from the set of semantic candidates Rk by use of the decision tree 82.
COPYRIGHT: (C)2008,JPO&INPIT
SUGAYA FUMIAKI
KDDI CORP