Topic Model Based Knowledge Graph for Entity Similarity Measuring

Entity similarity measuring is the basic work of academic search and recommendation. To analyze and measure the similarity among entities, existing methods are mainly based on either textual content or relationship unilaterally. Thus they can not measure similarity between paper and scholar or are limited in a small range of field. To address this, we propose a topic model, utilizing both textual content of papers and relationship between entities, and then introduce an algorithm for computing similarity between graph entities based on knowledge graph enhanced by topic model. Through the experiments we show that our model outperforms traditional topic model in topic coherence and we are able to list similar papers, scholars and conferences simultaneously and more accurately.