TRANSLATION MODEL LEARNING DEVICE, MACHINE TRANSLATOR, TRANSLATION MODEL GENERATION METHOD, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To solve a conventional problem of being unable to obtain suitable translation rules at high speed.SOLUTION: A translation model learning device includes: a parsing part for parsing one or more target language sentences to obtain one or more target language structures as information having one or more syntactic labels; a first translation rule generation part for obtaining one or more first translation rules that have the one or more target language structures and one or more original language structures of original language sentences, each corresponding to each of the target language structures; a segmentation part for segmenting the syntactic labels to obtain two or more extension syntactic labels; an aggregation part for aggregating the two or more extension syntactic labels into a predetermined number of groups so as to maximize an object function relating to a likelihood of the translation rules to acquire one or more aggregation syntactic labels; and a second translation rule generation part for using the one or more first translation rules and the one or more aggregation syntactic labels to acquire one or more second translation rules. The translation model learning device enables suitable translation rules to be obtained at high speed.