Lexical gap in English- Vietnamese machine translation: What to do?

Abstract

In English - Vietnamese machine translation (EVMT) project at Ho Chi Minh City University of Technology there are some problems that cause the system to malfunction. One of the most undesired phenomena is lexical gap. A lexical gap occurs in case of lacking Vietnamese equivalent word to English word. There are some approaches to this obstacle. Some researchers prefer replacing lexical gap by its nearest meaning word. However, in out project there isn't a conception database for Vietnamese word that helps to find acceptable word to lexical gap. Our approach for lexical gap establishes a word - to - phrase model for EVMT. Naive replacements of word by phrase keep semantic information as much as possible, however, lead translated sentences to syntactic mismatch, and therefore ungrammatical text. We restructure target sentence using phrase in dictionary as template of formal phrase structure in the model with some algorithms to keep target sentence be grammatical. This approach improves rule - based English Vietnamese machine translation. © 2010 IEEE.