| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a translation estimation device, method, and program capable of estimating translated words corresponding to keywords. <P>SOLUTION: An assembly of documents in a first language is searched according to keywords in the first language. The documents in the first language searched are machine translated to obtain documents in a second language where the keywords are masked as unknown words while words other than the unknown words are written in the second language. From a context containing the keywords masked as the unknown words in the documents in the second language, a keyword context vector is created. A representative context word is determined based on the keyword context vector and an assembly of documents in the second language is searched according to the representative context word. From a context in the documents in the second language searched, a second-language context vector having a certain number of words as elements is created. Based on a pair of the keyword context vector and the second-language context vector, words in the second language which correspond to the keywords in the first language are estimated. <P>COPYRIGHT: (C)2006,JPO&NCIPI |
