COMBINED TRANSLATION MODEL FORMING DEVICE, TEXT CLUSTERING DEVICE, AND METHODS AND PROGRAM THEREFOR

摘要

PROBLEM TO BE SOLVED: To form a high-accuracy translation model even when translated texts in a set of texts having a semantic correspondence are not sufficiently similar to each other, and enable proper clustering.

SOLUTION: A semantic correspondence replacement unit 2 forms a semantic correspondence replacement text group set 1' by mutually replacing semantic correspondences in a semantically corresponding text group set 1, a translation probability calculation unit 3 calculates word-to-word translation probabilities in the semantically corresponding text group set 1 and the semantic correspondence replacement text group set 1', to form a correspondence non-replacement translation model 4 and a correspondence replacement translation model 5. A translation probability composition unit 6 calculates a translation probability that each word in the semantically corresponding text group set 1 is translated to the other word from the translation models 4 and 5 to form a combined translation model 7. The combined translation model 7 is used to cluster a group of input texts.

COPYRIGHT: (C)2011,JPO&INPIT