Lexical Knowledge Extraction From Machine Readable Dictionary

Lexical information will be an essential part of natural language processing systems, its scale and quality determines whether it will be successful, which is an common agreement among the researchers.The construction between conceptions is an important research area in the construction of ontology, especially Chinese ontology. They are the base knowledge that would be useful for the further development of natural language processing. It is not only the theory base of making linguistic knowledgebase, but also has a wide application.And in a certain degree the coupling relations between conceptions represent the semantic relations between words. So the automatic extraction and construction will be able to provide an ideal platform for the development of computational linguistics.At present the construction of semantic relations still rely on the manual method, the workload is too high. If we can import the aid of computer, then the efficiency will be improved greatly. We believe that the relations between words have their own internal law, which would be helpful to the automatically construction by programs.This paper uses a machine readable dictionary as the resource. First my research is on the extraction of conceptual relations; the method is combined with rules and features. There are many potential relations embedded in the machine readable dictionary. I tag some nouns'paraphrase as training corpus; extract the useful patterns from them to construct the semantic relation between nouns and their paraphrases. Then use the features from context to build corresponding statistics model for the disambiguation. The implement of this model turns out to be a good result after applied to Chinese Dictionary for Application.Then I made some attempts on the automatic extraction of intentional conceptual meanings, mainly on the research of feature of nouns and some of adjectives. we divide the words into several groups, then use a set of extraction patterns to locate specific knowledge and relations among different knowledge automatically, at last a supervised machine learning method is proposed in this paper, which is applied in result filtering. By this way some experimental results are successfully gained.