A Post-Search System for Grouping Relevant Academic Articles into Research Topics

Abstract

In this paper, we develop a smart topics finding system to organize the research topics into a hierarchical tree. First of all, we use some natural language processing techniques to convert the collected snippets into a series of meaningful candidate terms. Second, we use a suffix tree clustering with threshold and two-steps hash to generate the topic label. Third, we use a divisive hierarchical clustering method to arrange the topic label into a hierarchical tree. In this paper, we use precision and normalized Google distance to measure the quality of topic results. According to the results of experiments, we conclude that using our system can give significant performance gains than current academic systems.