Derivational morphology to the rescue: How it can help resolve unfound words in MT

摘要: Machine Translation (MT) systems that process unrestricted text should be able to deal with words are not found in the MT lexicon. Without some kind of recognition, parse may incomplete, there is no transfer for unfound word, and tests transfers surrounding will often fail, resulting poor translation. Interestingly, much has been published on unfound-word guessing context although such work going other applications. In our IBM system, we implemented a far-reaching strategy recognizing based rules word formation generating transfers. What distinguishes approach from others use semantic syntactic features both analysis transfer, scoring system assign levels confidence possible structures, creation transformation component. We also successfully applied derivational morphological non-derived words.