A Study of Sentence Ecdysis in Chinese-English MT
LI Ying (signal & information processing)
Directed by HUANG Zeng-yang
Sentence Ecdysis (SE), as a common structure in sentences, exists widely in natural languages and is not processed properly in Machine Translation (MT) systems at present. Based on Hierarchical Networks of Concepts (HNC) theory, this dissertation puts forward some rules and algorithms for processing SE. These rules and algorithms are as follow:
Ø The rules and algorithms for processing Constituent SE (CSE);
Ø The rules and algorithms for processing Prototype SE (PSE);
Ø The rules and algorithms for processing Wrapped SE (WSE);
Ø The algorithm for disambiguating the structure of SE.
These rules and algorithms mentioned above make it possible for computer to process SE, which marks the transformation from description of SE to calculation of SE and promotes the development of HNC-based MT Engine.
Besides, this dissertation sheds light on SE and relative subjects from different viewpoint, such as the follows:
Ø Expatiating on SE in theory;
Ø Comparative analysis of SE between Chinese and English;
Ø Mathematical description for sentence category space;
Ø Formalization of HNC-based Interactive Engine
This dissertation includes 6 parts. In the first two parts, the review of MT and summary of HNC are given, then comes the main contents of this dissertation. In the third and fourth parts, some linguistic descriptions of SE, some definitions of sentence format and comparative analysis between Chines and English are given first. Then, the dissertation focuses on the questions of how to process SE in Chinese-English MT, and gives some rules and algorithms for processing SE. In the last two parts, some tests of these rules and algorithms in the current MT systems and statistical analysis in large-scale real texts have been taken to manifest the importance of SE processing. Besides, some conclusions drawn from this dissertation are given and the futher works are introduced.
Key words: Machine Translation£» Hierachical Networks of Concepts£» Sentence Category£» Semantic Chunk£» Sentence Ecdysis