Domain Experts Identification Based On Dynamic Topic Model

With the development of the Internet,the scale of information has increased at an alarming rate.As human society rapidly moves from the information society to the knowledge society,experts with specialized knowledge have become more and more valuable resources in the modern society.Experts finding,as an important aspect of the entity search,aims at finding experts with expertise in a particular domain based on specific needs.Expert identification is an important and significant field of study,and many researchers are devoted to this issue.The main content of this paper is the identification of domain experts based on dynamic topic model.It is divided into three parts.Firstly,a series of tasks are carried out by the traditional topic model.According to the structural characteristics of social network data,a system framework for domain expert identification in social networks is designed.The method takes full account of the semantic information of the relevant corpus of experts and proposes a method based on the interest to build an expert network.The method of link analysis is used to analyze the authority of experts.Is a way to think about the full combination of semantics and network structure.The method is applied to the real data set of quizzes and communities and compared with other methods.The results confirm the validity and advantage of this method.Then we move on with the dynamic topic model related research.The dynamic topic model is basically applied for the evolution of the literature.Combined with the characteristics of large time span and meaningful time series of the existing corpus,we choose the dynamic topic model to identify domain experts and opens up a new research field of dynamic topic models.Combined with the methods of expert modeling based on document modeling and contour modeling,the temporal features of plain text data are utilized while considering the semantic information,and the author's authority is accumulated according to certain rules.This method is compared with the method of using the traditional topic model,the experimental results have better accuracy.On the basis of the second part of the work we carried out the research of expert team formation.Conform to the trend of network through the integration of multi-source data approach to the relationship between experts and the use of community division ideas generated expert team.This method avoids the arduous task of defining and calculating communication overhead while ensuring lower communication costs.Through the case and experimental data can prove the effectiveness of this method.This topic uses the topic model to carry on the effective research to the domain expert's recognition in each kind of situation,and takes the expert team's recognition research as the expansion,verifies the feasibility of each method through the experiments and the case analysis.