| 摘要 |
<p>PURPOSE: To enable a user or the system itself to correct the correspondence relation requested by the system between an original text and its translation by using case information of one language which is estimated to be correct as case information for extracting a translation rule. CONSTITUTION: An input part 1, a display part 2, and a storage part 3 are placed under the control of a control part 4, which performs a process for learning a translation rule. The dependency structure of an original text side is collated with the dependency structure of its translation side to generate case frame correspondence candidates. The generated case frame correspondence candidates are stored in a case frame correspondence buffer. Further, when there is a case frame correspondence candidate for which a translation rule is not extracted and the declinable word of the case frame meets conditions of parts of speech and words, a case frame correction routine is called and a translation rule extraction routine is also called to generate a user translation rule from the case frame, thereby outputting a user translation rule in a translation rule registration buffer to a user rule dictionary.</p> |
