The Construction Method Of Knowledge Graph In Micro-blog

With the development of information technology, and the popularity of PC and mobile terminal, more and more people began to contact the internet. Various social platforms, such as face book、 sina Micro-blog 、tencent Micro-blog and so on, had greatly changed our way of life, and the information communication online was becoming more and more frequent and convenient. Now, the internet had become a global information resource database, which contained a large amount of information. With the characters of massive,complex and changeful, big data brought great difficulties for information retrieval. In order to improve the quality of retrieval and realize human-web interaction, also achieve the prediction of human search purpose finally; knowledge graph was born in this case.With the development of social network, social search had become increasingly important. Construction knowledge graph in social networking platform is crucial for improving the quality of social search about massive data in network platform. In this paper,we constructed knowledge graph about sina Micro-blog, and all of those works were on the basis of that concept and relations among concept were extracted.(1), this paper defined the notion of equilibrium based on triadic closure to compute the weight of relationships firstly.(2), and then improved PFNET was used to generate knowledge graph about a single topic,we called CKG(Compact knowledge graph).(3), for the same topic, a serious of CKGs were generated in different time, also some of the concepts were the same, and those CKGs were merged based on same concept, which formed a single knowledge graph.(4), when it came to CKG merging in different topic, the concepts could be extracted firstly, if there were common concepts among different knowledge graph, CKG could be merged through those concept; if there were no common concepts, we used relations among concepts to merge, and this work divided into two case: only one relations between CKGs and multi-relationships among CKGs,also each cases could be treated separately. Specially, in the case of multi-relationships, we mainly calculated the betweenness values for each relationship, and only retained the edge whose betweenness was maximum.In this paper, we verified this method by experiments, and a knowledge graph was effectively formed in sina Mirco-blog. So our work had certain and practical significance for the development of knowledge graph.