Diversified Portrait And Retrieval Of Social Relations In Social Media

The research of social relations has always been an important topic in the field of sociology.With the rapid development of the Internet technology and social media platforms,more and more people are active in social media.This provides a new way for the research of social relations.People's interaction in social media is the mapping of real social network on the network,which provides reliability for researching social relations through social media.Moreover,massive social media data also brings great convenience to the research of social relations.This paper aims to make diversified social relation portraits through social media data.Diversification is mainly reflected in the diversification of data and methods.The diversification of data means that we respectively make portraits on the speech data of people and the speech data of third parties.And the diversification of methods means that we respectively make non-intuitive social relation semantic portraits based on representation learning and intuitive social relation semantic portraits based on topic analysis,besides,this paper proposes the portrait methods of social relation intensity based on different social relation semantic portrait methods and the theory of social relation intensity.Furthermore,we design and develop an online social relation automatic portrait tool to visualize the portrait results,and provide functions such as relation classification,node merging,core character positioning and relation retrieval.Finally,based on the Enron email data set and Sina Weibo data set,we carry out the comparison experiments among different portrait methods,including the comparison experiment of different portrait methods on relation semantic representation and the comparison experiment of different portrait methods on relation intensity representation.We verify and analyze the consistency and difference of relation semantics and relation intensity represented by different portrait methods.