Semantic Framework Filling Research Based On Information Extraction

Nowadays, with the rapid development of Internet technology, the electronic form of document resources is dramatically growing exponentially. Faced with such a large amount of electronic information, the problems of how to extract the key information from the vast amount of information needed and structuring of unstructured data become especially prominent. In order to solve these problems, this paper proposes a study of semantic frame filling technique based on information extraction, which is of great research significance and use value.This thesis focuses on the main parts as follows:1. A deep study of the extraction rules of information extraction and in view of the existing shortcomings of traditional research. Methods, which gives a kind of text information extraction algorithm based on the "HowNet".The description of the concept of entity in accordance with the "HowNet" and the similarity comparison of "primitive" make the similarity comparison of the word and sentence achieved.On this basis, the key information of text extraction is realized.2. According to the extraction results of non-structural problems reflected in traditional information extraction methods and combining with frame, this paper proposes a semantic framework filling method based on key information. This method can make use of semantic framework to represent knowledge and make unstructured text shifted into structured text representation.3. When matching semantic slots of semantic framework for filling key information, it gives a matching algorithm of semantic similarity based on slot. The similarity comparison of feature words and key sentences to corresponding semantic slot makes matching and filling of semantic slot on key information.4. In the processing of mining missing semantic slot information in the semantic framework, this paper gives the reasoning algorithm based on rules. Reasoning method based on the description logic has limitation in the expression and efficiency. That inference mechanisms based on rules are added in the process of reasoning can shield all inference rules of contradiction and increase the expressive ability of the system and guarantee the consistency reasoning.Finally, by the comparison of traditional information extraction, information extraction efficiency is higher for a method from this paper.At the same time, it also testifies the effectiveness of the subject and makes a foundation of further studies in the field of information extraction.