Expert Identification Based on Dynamic LDA Topic Model
Renjun Chi, Bin Wu, Lin Wang · 2018
In recent years, human society is transferring from information society to knowledge society. Experts mastering professional knowledge are becoming more and more valuable resources in the society, therefore Expert Identification, also known as Expert Finding, became an important research field. Existed Expert Identification work is mainly based on traditional information retrieval, or standard topic models. Experts finding still faces a lot of problems, such as the missing of semantic information or the inaccuracy without changes over time taken into consideration. This paper presents a domain Expert Identification method with the improved dynamic LDA algorithm which solves these shortcomings of existed methods. Based on the standard LDA model, this method divides the corpus with large time span according to time to apply the dynamic LDA model and combines profile modelling and file modelling for expert modelling. In addition, this method considers both the semantic information of the domain and expert authority. Experiments show its feasibility and effectiveness, and its advantage over the traditional static topic model. It has opened up new application fields of dynamic topic model.