摘要 |
Words that are considered inappropriate as translated words in a specific field are efficiently detected. At a client (10), a translation instruction unit (11) instructs the translation of an original document by a translation engine (20). A morphological analysis unit (13) applies morphological analysis to a translated document received from the translation engine (20) to extract compound words. A search instruction unit (15) inputs a search query that includes the extracted compound words to the search engine (30) to find (a) the number of perfectly matching pages on a specific site, (ß) the number of co-occurring pages on the specific site, (?) the number of perfectly matching pages on all sites, and (d) the number of co-occurring pages on all the sites. Then, an evaluation processing unit (17) compares a or ß with ? or d to classify the compound words into levels that indicate whether appropriate as translated words in a specific site, and a dictionary updating unit (19) updates the dictionary in the translation engine (20) in accordance with instructions given by a user based on the display of said levels. |