Research On Extracting Term Relationships Based On Semantic Grammar

Knowledge representation and acquisition is an important part of artificial intelligence. And the acquisition of term relationship is a major research direction of knowledge extraction. Terms are words which refer to entities. Relationships between terms refer to associations between terms. The acquisition of term relationship refers to the process which separates the relationships of terms from the structured or half-structured texts and stores them into computer in some certain form.Although term relationships might be similar of definition, they are quite different from each other when they are used in different situations. Term relationships are hidden in a huge number of Internet information, so the manual acquisition of them will consume a large amount of manpower and material resource. Meanwhile, even if relationships are extracted automatically, to verify them is another difficulty task.This thesis introduces a semantic grammar based on approach to extracting term relationships from text, which has achieveed four major results as follows:(1) A detailed characterization term relationshipsRelationships between various terms are summarized and characterized. The characterization of a term relationship includes formal definition, predicate representation by describing the types of object and nature of relationship, and factors of semantic restriction in the relationship.(2) A method of extracting term relationships from textEach relation grammar set is summarized by analyzing a few text sentences. The grammar is used by OMKast to parse a Web corpus which possibly contain term relationships. At the same time, according to the obtained results, the solution to problems about obtained results is put forward to guide the grammar construction process to improve the correctness and accuracy rates by analyzing the problems in relationship extraction.(3) A method of designing axioms of term relationshipsAccording to the relationship definition, from the angle of symmetry, transfer, implication, the intersection, all the relationship axioms are obtained to construct axiom system. Take use of LL1 grammar to validate syntax correctness. Take use of the formula deduced form predicate logic to validate consistency of the (or contradiction). Refer to the Suwat concept map and semantic network to respectively inspect axiom vacancy and relationship vacancy and give the corresponding algorithms.(4)The verification of the term relationshipFirstly, it is pretreatment to strip away the noise words in the relationship by the establishment of stripping word lexicon and the stripping sentence model. Then analysis and summarize semantic features, the term relationship grammatical features and statistical features to validate relationship relative correctness. Finally, discuss the problem on sequence and on the efficiency of searching to validate relationship relative correctness by combining with the term relation related axioms. Then the corresponding inspection algorithm is given.