Research Hotspots Evolving Action Detection based on Time Sequence Journal Topic Model

Abstract

Since the research hotspot development in academic fields is mainly reflected through academic journal contents, how to analyze the evolving action of academic journal related topics is a huge factor for researchers in grasping the tendency of research hotspots. This paper considered and combined two characteristics of academic journals: 1) topic property and 2) time-sequence feature to realize journals’ time-sequence topic extraction, which also puts forward the TS-JTM (Time Sequence Journal Topic Model) at the same time. On the basis of TS-JTM, we developed topic-snapshot journal hotspot evolution model based on time sequence, and proposed a method which could detect the continuing, emerging, splitting, amalgamating or disappearing between two neighbor topic-snapshots, with adopting topic similarity measurement based on Kullback-Leibler (KL) Divergence. Our experiments show that the proposed method could realize evolving analysis of journals’ research hotspots effectively.