Chronological Scientific Information Recommendation via Supervised Dynamic Topic Modeling

Scientific information recommendation is crucial to assist scholars for their researches. Citation recommendation is an important field of scientific recommendation. Traditional approaches ignore the chronological nature of the citation recommendation task. In this study, I propose the "Chronological Citation Recommendation," which assumes initial user information need could shift while they are looking for the papers in different time slices. Specifically, I employed a supervised dynamic topic model to characterize the content "time-varying" dynamics and constructed a novel heterogeneous graph that contains dynamic topic-based information, time-decay citation information and word-based information. I applied different meta-paths for different ranking hypotheses, which carried different types of information for citation recommendation in different time slices along with information need shifting. I plan to generate the final "Chronological Citation Recommendation" rankings by feature integration using Learning to Rank. "Chronological Citation Recommendation" will recommend time-series ranking lists based on initial user textual information need. Preliminary experiments on the ACM corpus show that chronological citation recommendation will significantly improve the citation recommendation performance.