| 摘要 |
PURPOSE:To provide a natural language processor which can provide a proper translated result even in an unregistered word by discriminating whether the unregistered word can be divided into plural words registered on a dictionary or not when the unregistered word is present on an input character string and replacing the unregistered word with the plural words when it can be divided. CONSTITUTION:A language analysis part is composed of a translation processor 2 and a dictionary 4 for translation. For a source sentence inputted by an input device 1, morpheme analysis is performed at the input character string analyzing means of the translation processor 2 by using an analysis dictionary 4a, and it is decomposed into units to perform processing in the source word. After wards, the structure transforming means of the device 2 transforms the structure of the source sentence provided by syntax analysis into the translated word or the structure of the translated sentence by using a translation dictionary 4b. In such a case, the analyzing means for morpheme analysis is provided with a means for normally referring to the dictionary, a means for dividing the unregistered character string into plural partial character strings, and a means for referring to the dictionary again concerning the character string provided by the division. |
