| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To enable selection of a translation candidate that can optimize a desired translation evaluation scale using the Bayes risk minimizing reference from translation candidates obtained from a plurality of translation devices. <P>SOLUTION: An optimal translation sentence selection device comprises: a second translation control section 48 that acquires a plurality of translation sentence candidates for an inputted translation source language sentence from a plurality of translation devices 43A-43C; a second feature quantity calculation section 49 that, for each of the acquired translation sentence candidates, calculates a plurality of feature quantities correlated to a predetermined translation evaluation scale; a translation sentence score calculation section 50 that, for each of the plurality of translation sentence candidates, calculates a translation sentence score using a learned optimal translation sentence selection model; and an optimal translation sentence selection section 51 that selects an optimal translation sentence from the plurality of the translation sentence candidates. <P>COPYRIGHT: (C)2013,JPO&INPIT</p> |
