Improving Information Extraction Using Knowledge Model

Improving Information Extraction using Knowledge Model Yue Chen School of Computer Science and Technology, Beijing Institute of Technology, No.5, Zhongguancun Nandajie Haidian District, Beijing 100081, P. R. China chenyue@bit.edu.cn Abstract. This paper describes the work on automated Information Extraction that accepts arbitrary text and extracts information from the text. A new approach to implement Information Extraction system is proposed in this paper. Firstly, the article will be decomposed according to paragraph, sentence and phrase. Every sentence will be compared with the knowledge node, and then append the information extracted to the knowledge model. Finally, the answers are generated to the questions about the input text. With the experimental corpus the accuracy rate of knowledge matching is 63.5%, and accuracy rate of question answering is 65.0% with the system knowledge model. Keywords: Natural Language Processing, Information Extraction, knowledge model, knowledge representation. 1 Introduction This paper proposes an approach to establish an Information Extraction (IE) system by using knowledge model obtained from articles. A simple knowledge representation method, where knowledge is stored in a graph structure, is adopted for this system. IE system intends to test the ability of computer to extract the information from an article, and the main method of testing accuracy is to generate the answers automatically with the input text and questions. The objects of extraction in sentence include attributes, relationships and events. In the field of Natural Language Processing (NLP), the theory of knowledge model is conducive to a number of research directions, and many areas in NLP need to access the knowledge from article, which make the study of IE system has its important significance. Questions can be formalized to query that related to the knowledge structure using the technology of knowledge model in QA system (McKeown, 1979). Machine translation (MT) system will benefit from knowledge model because it can generate more easily translated sentences for the input articles ( Mitamura and Nyberg, 2001). IE System mainly use template method, while the knowledge model maintains the knowledge node which includes the characteristics of knowledge. Knowledge model can improve the accuracy and robustness of IE system (Shinyama et al., 2002). In information Retrieval (IR), the document stored in database can be changed into knowledge structure with knowledge model technology so that improve the performance as well as precision of search result (Zukerman and Raskutti, 2002). In Summarization, the center of an article can be extract effectively by understanding the knowledge of article with knowledge model. It can extract a summary from the chapter, paragraph, or sentence level for different needs, and rely on natural language generation to get a more accurate digest (McKeown et al., 2002). In paraphrasing, since that the process of paraphrasing can be considered as translation between the same languages, the part of natural language generation in knowledge node will improve the accuracy of paraphrasing through making the sentences and articles into knowledge representation (Iordanskaja et al., 2002). PACLIC 24 Proceedings 613 2 Related Work The research of knowledge model technology originated in the …

Content truncated to 500 words