Exploiting Temporal Authors Interests via Temporal-Author-Topic Modeling

This paper addresses the problem of discovering temporal authors interests. Traditionally some approaches used stylistic features or graph connectivity and ignored semantics-based intrinsic structure of words present between documents, while previous topic modeling approaches considered semantics without time factor, which is against the spirits of writing. We present Temporal-Author-Topic (TAT) approach which can simultaneously model authors interests and time of documents. In TAT mixture distribution over topics is influenced by both co-occurrences of words and timestamps of the documents. Consequently, topics occurrence and correlations change over time, while the meaning of particular topic almost remains unchanged. By using proposed approach we can discover topically related authors for different time periods and show how authors interests and relationships change over time. Experimental results on research papers dataset show the effectiveness of proposed approach and dominance over Author-Topic (AT) model, due to not changing the meaning of particular topic overtime.