On Halliday’s Thesaurus Method

In translation, one of the most important research fields is machine translation. Halliday had had remarkable and outstanding work in this field, which played a critical role in the history of machine translation. In all the work he has done, nothing is more influential than the work of thesaurus method he proposed in his early academic career. The thesaurus method is rooted in systemic functional grammar, so through this research one of the most crucial research path of combining systemic functional grammar with statistics and computer science was created.This thesis studies Halliday’s thesaurus method, a method to help the machine to translate with thesaurus. This method is not antagonistic to rule-based method and statistic-based method, but they counterpose to each other. This research mainly contains: first, the linguistic basis of Halliday’s thesaurus method and methods used in machine translation; second, the statistical methods and computational technology that are used in the proving or disproving of Halliday’s thesaurus method; third, the application and developing prospects of thesaurus method in machine translation systems. The aims of this research lie in exploring Halliday’s thoughts in computational and quantitative studies, and in making further efforts in investigating the application potential of his theoretical methods in modern machine translation systems. Among them, the emphasis is to evaluate its application value in statistical machine translation.This thesis mainly uses the following two methods: literature research and the method of computational science. Through literature research, a preliminary consolidation of Halliday’s study achievements in the area of machine translation and mechanical dictionary in the years of 1950 s and 1960 s is done, especially that of “thesaurus method” which is of great historical significance as well as practical value. The method of computational science refers to the two small-sized English-Chinese parallel corpora that are built during the research process, based on which a test was given, attempting to check the feasibility of this method in applying to machine translation. Improvement has been made to this method, and the direction of further research was offered.Through the above theoretical discussion and experimental research, three important understandings are established: 1) From the perspective of linguistic theory and the development history of Systemic Functional Grammar, thesaurus method has organically developed Saussure’s theory of symbolic system, and further intensified the choice in paradigmatic relations, thus promoting the development of Scale and Category Grammar to that of Systemic Grammar; 2) Translation equivalence plays an important role in machine translation, and affects the quality of translation greatly, thus, the development of thesaurus method relies on a further development of equivalence theory, which deserves in-depth research; 3) From the perspective of the development of machine translation, thesaurus method was applied successfully to the first generation of machine translation system, and has great application potential in the third generation of this system, that is the machine translation system based on corpora. Compared with the current commercial translation software like Kingsoft and Google, this method has better outcomes in the aspect of words/phrase choosing and translating. Still, this research requires more close integration with statistics and computational science.Further research direction lies in three aspects. First, the lack of existing English–Chinese translation probability table has brought some difficulties to this thesis; therefore, we will strive for building larger corpora, making experiments more convincing. Second, no automatic translation or contrast on time with other translation software was realized; therefore, this method should receive further examination from experts of machine translation. Furthermore, the thoughts of equivalence in Halliday’s thesaurus method are to be studied in depth.