Research On Personal Relation Extraction Method For Social Network Application

Personal relationextraction is a task that automatically recognizes personal relation from unstructured or semi-structured documents, and stored them in a structured format.In this big data era of information explosion, it is necessary to find diverseof relationships between the personsrapidly and accuratelyutilized the technology of personal relation extraction, to lay the foundation for the application and study of social network and other related fields.In this paper, we studied the personal relation extraction method based on thestacked denoisingautoencoders,and further built a social network. The main research contents are summarized as follows:Firstly, we studied the information acquisitiontechnology based on the Internet and the automatic generationtechnologyofrelation instance datasets based on hudongbaike.The design method of topic crawler system for personal relation extraction application and the name recognition method combined LTP and NLPIR and further confirmed by hudongbaike and the automatic annotation method of relation instance based on hudongbaike were proposed.The purpose is to provide corpus for the parameters learning and performance testing of personal relation extraction model.Secondly, we explored the personal relation extraction method based the stacked denoisingautoencoders.For the situation that sentence-level personal relation extraction leads to highly sparse eigenvectors due to the number of feature word is too small, we provided a feature word expansionmethod based onsynonymCilin to ease the phenomenon.We also explored the contributions of the dependency parser informationand the semantic information for personal relation extraction. Respectively, we explored the effects of personal relation recognition based on the shallow layer denoisingautoencoders and the deep layer stacked denoisingautoencoders.Finally, we designed a prototype system for personal relation extraction and studied the building method of social network based on personal relation extraction technology.The prototype system containstopic web crawler, web content extraction, text analysis and processing, personal relation extraction, social network building, realizing complete process of building social network based on personal relation extraction technology.The building method of social network based on personal relation extraction technology presented in this thesis is suitable for utilizing the large-scale data of web to build social network and lays the foundation for the further application and research of social network.