Chinese-english Machine Translation System English Generated In A Choice Of Words Model

This thesis presents a description of a lexical selection model applied in English generation of the Chinese-English machine translation system. This model realizes lexical selection using semantics knowledge of lexical collocations. It combines two lexical selection methods,that is,the method basing on semantics patterns and the method basing on examples. The former is used to perform primary lexical selection. And the later,on the other hand,is used to do further lexical selection. The semantics patterns used in this system are named fuzzy semantics patterns. They are different from the traditional ones and will be trained from an example base with a training algorithm. The examples used for comparison will be extracted from a corpus called SEMCOR. SEMCOR is semantically tagged with WordNet senses. However,the semantics information of the examples is required to be HowNet concept according to the interlingua used in this system. Therefore,before the extraction,the corpus is transformed with an algorithm that maps WordNet senses to HowNet concepts.The whole thesis is made up of the following five chapters:Chapter One is the preface. It shows the importance of studying the problems of generation module in interlingua-based machine translation system. And then it discusses the problems we concerns - lexical selection in target language generation module. At last this chapter gives the topic of the thesis.Chapter Two puts forward the general idea about the lexical selection model and then gives a brief introduce to the interlingua and Hownet,which will be used in succedent discussions.Chapter Three introduces the construction of example base and semantics pattern base,which are the main data resources of the lection selection model. In this chapter,the automatic algorithm for mapping WordNet senses to HowNet concepts and the algorithm used to acquire fuzzy semantic patterns from examples are both described in detail.Chapter Four describes the algorithms of the lexical selection model,including the algorithm basing on fuzzy semantics patterns and the algorithm basing on examples. At last it also brings forward a brief discussion about counter semantics pattern and counter example.Chapter Five is the concluding remark. It summarizes the advantage and deficiency of the lexical selection model,and then gives some ideas of further improvement.