A Study On Ambiguities And Disambiguation In English-Chinese Machine Translation

Machine Translation (MT) has not achieved any real breakthroughs in its outputquality despite of more than sixty years of development. Ambiguity is still a majorproblem in its translational products. Since ambiguity typologies vary greatly accordingto different standards, and for the convenience of analysis in this study, the “ambiguity”in this thesis has been classified to lexical ambiguity, syntactic ambiguity, semanticambiguity and pragmatic ambiguity. Therefore, one purpose of this thesis is to exploreambiguity issues from a purely linguistic perspective, aiming to provide a generaloverview about the different kinds of ambiguities both in the English and Chinesedomains in English-Chinese (E-C) MT.Ambiguity is a complicated phenomenon in natural languages and is caused bymany reasons. Anyone who wants to use one general schema or rule to disambiguatethis issue is doomed to fail. In this study, we consider one of the main purposes of MTis to offer the user or receiver information. Therefore, Skopostheorie (i.e. the overalltranslating purpose is the most important factor in translation process) is adopted as thetheoretical foundation, and “adequacy or appropriateness rather than equivalence orexactness” is set as the guiding principle for disambiguation in this paper. Based onsuch framework, we focus on the similarities between English and Chinese ambiguitiesto primarily discuss word-sense disambiguation and syntactic disambiguation.Specifically, for word-sense disambiguation, we have attempted three methods, i.e.,corpora-based strategy, knowledge-based strategy and statistics-based strategy; then foreliminating ambiguities at syntactic level, we have explored rule-based strategy andhybrid approach.Around the main theme of ambiguities and disambiguation in E-C MT, data arecollected through a hybrid method of qualitative study and sampling strategy. Thequalitative study aims to answer what kinds of ambiguities exist in Chinese and Englishtexts and what resolutions can be used for disambiguation in their MT. The sampling strategy is used to offer representative samples for the analysis of ambiguity types anddisambiguation resolutions.Through this study, we found that ambiguity is a very complicated phenomenonand exists commonly both in English and Chinese. However, E-C MT systems still havemany limitations to deal with this issue satisfactorily. In terms of the complex nature ofambiguity, the road of disambiguation is doomed hard and long. In E-C MT, all in all,whether for English ambiguities and disambiguation, or Chinese ambiguities anddisambiguation, both need more studies, and require much additional time and work.