Research On Key Technology And Algorithms In Unstructured Information Extraction Based On Cognition

With the rapid development of the computer technology, the information extraction technologyhas become one of the hot topics of natural language processing field in recent years. Furthermore,machine learning, text mining and graph algorithms have already been applied to informationextraction. However, the performance of information extraction algorithms can not be satisfied andthere are many challenging problems for further research. In this paper, by analyzing the drawbacksof the existed document representation model, we apply graph model, conditional random fields’theory, machine learning-relative knowledge to implement information extraction algorithms. Inorder to improve information extraction’s performance, some information extraction algorithms areproposed in this paper, such as named entity recognition algorithm based on rules, an improvedperson’s name recognition algorithm based on rules, named entity recognition algorithm based onrules and conditional random fields, Chinese organization’ abbreviation name generation andrecognition algorithm based on rules, person’s relationship recognition algorithm based on textclassifiction. Furthermore, the effectiveness and efficiency of these algorithms are all validated byexperiments. These proposed algorithms in this thesis broad the prospect for the development ofinformation extraction technology.