Research Of Topic Focus Recognition And Popularity Evaluation In Online Social Network

Online social network has become an important platform for users to share information and express opinions.When users post comments concerning certain themes or discuss with other users on a social network,then an online social network topic will be formed.Due to the differences in cognition of various users,the perspective of viewing topic and the concerns of comments are also different,therefore online social network topic will contain some different focus.By studying topic focus on online social network,it can help managers grasp user demands,track the development of user concerns,and provide theoretical support for brand management and crisis management.Microblog and forums representing online social networks are two important platforms for social network topic.Since the topic corpus of the aforementioned two different situations is mainly composed of short texts,this thesis first studies the corresponding recognition methods of topic focuses based on the two different data characteristics of Weibo topics and sub-forums.It is foundation of online social network topic focuses research.Secondly,by modeling the interaction of the focuses,we study a model of evaluating topic focuses popularity and describe the popularity trend of the focus.According to the popularity function of the focus,we analyze the suitable intervention time of different types of focuses.The detailed research work and innovation are as follows:(1)A Weibo topic-oriented focus recognition method.Based on the similarity of the texts posted by "neighboring" users in the Weibo topic corpus,this thesis studies a short text merging strategy.Assigning different weights to short text according to its contribution in the virtual long text,we construct a mixed-weighted Weibo topic focuses recognition model.The model only uses the information of the Weibo topic corpus itself without other external information.At the same time,the weighting strategy can reduce the influence of noisy texts,solve the sparsity problem caused by the fewer words co-occurrence of short text,and achieve the goal that focuses are recognized effectively from Weibo short texts topic corpus.The proposed model conditionally relaxes the assumption of independence between the texts in LDA model,further extends theories of topic models,and also enriches the research content of topic models in the field of topic recognition.(2)A sub-forum-oriented focus recognition method.Due to the divergence and innovation of user reply posts in the sub-forum corpus,the focuses distribution of some main posts is not exactly the same as that of reply posts.According to the characteristics of weak relevance between posts,this thesis studies a focus recognition method based on multiple-iteration merging strategies.The model obtains the focuses distribution of short text posts according to each iteration,and re-clusters and merges short text posts with high similarity of focuses distributions.After multiple iterations,texts in the final cluster merges the virtual long text.The consistency of the short text focuses distribution is high in each virtual long text,and the increase in the co-occurrence of correct words reduces the sparsity.Experiments show that the proposed model can better recognize focuses in two data sets,and the focuses' coherence is also better than the comparison model.The sensitivity analysis of related parameters verifies the effectiveness of the new method.(3)The popularity evaluation model and intervention analysis of topic focuses in online social networks.According to the interactive influence between topic focuses,this thesis studies a focus popularity evaluation model based on Aging theory,which describes the popularity trend of the focus by establishing the popularity function of the focus.The model measures the interaction of the focuses,redesigning the nutrients into two categories at each stage,which can improve the effect of modeling the focuses popularity evaluation and divide the focuses into two types according to the ratio of the two types of nutrients in focuses.By calculating the first derivative of the focus popularity function,the trend of the growth rate of the popularity function is studied.And by considering the types of focus,the suitable intervention time and intervention method of each focus are analyzed,which expands the content in the field of information dissemination and information intervention.Finally,this thesis takes 4 data sets as examples to verify the effect of the proposed model describing the popularity trend of the two types of focus,explains the reasons for the popularity change of the focus through the change of nutrients,and finds the corresponding focus intervention time according to the popularity growth rate.This thesis takes online social network topic focuses as the research object,extends the research about social network topics,and enriches the methods of short text topic recognition.At the same time,the research results also provide theory support for managers on understanding user concerns,developing corresponding marketing strategies and public opinion management.