The English Syntactic Analysis system for the MATE-EK (MAchine Translation Environment - English to Korean) is designed and implemented in this thesis. Input of the system is a initial annotated syntactic tree, the output of the English Morphological Analysis System. And, output is a English dependency structure, the input of the English-Korean Transfer System.
The system consists of phrase-structure constructing phase and dependency-structure constructing phase. The former consists of unit phrase processing, coordination processing, sentence category processing, and phrase structuring modules. The latter consists of nominalization and dependency structuring modules. Each module is described in GWL-II(Grammar Writing Language-II) and converted to LISP program to execute.
Mixed-mode parsing method is adopted to parse English sentences. Top-down parsing method is used to construct unit phrase, and bottom-up parsing method is used to connect unit phrases. Lexical ambiguities are handled by controlled backtracking using the guessing information of the English Morphological Analysis System. also, syntactic ambiguities are handled by controlled backtracking using deep-case information.
Comparison with other systems shows that the system has good processing efficiency and broad parsing domain.