| As an important branch of Natural Language Processing (NLP) in computer, the research of machine translation has significant social and economic values. The Example-based Machine Translation (EBMT) is an empirical approach, which avoids complicated deep-level linguistic and semantic analyses, and provides a new way for machine translation.In this thesis we review the history of the development of machine translation, and describe the status of machine translation in detail. We also introduce the main theories and techniques in the machine translation field nowadays, especially the Corpus-based Machine Translation (CBMT) in'90s, compare advantages and disadvantages between the new empirical approach and the old rational approach, and then we discuss the development trend and application prospects of machine translation.This thesis emphasizes on the EBMT methods based on bilingual aligned trees. The use of bilingual aligned trees improves the adaptability and accuracy of EBMT systems. Tree-alignment pays attention to the characteristics both of word-alignment and structure-alignment. The structure of tree can effectively contain the structural and syntax information in sentences. This thesis introduces several important EBMT models based on bilingual aligned trees.And then we present the tagging sets and rules of the EBMT system, the way of alignment and the expression of link relations. The EBMT system uses Extensible Markup Language (XML) to represent the bilingual aligned tree-bank and its storage structure. Meanwhile, we also present the translating algorithm of the EBMT system.In this thesis we implement an EBMT system based on bilingual aligned tree-bank by Visual C++, and use the MSXML parser of Microsoft to operate the XML document of example sentence corpus. Then we introduce some problems and possible solutions in implementation process.Finally we do some translation system testing, summarize the strongpoint of this EBMT system, and discuss the prospect of EBMT in future. |
PDF Full Text Request