Citation Recommendation via Time-series Scholarly Topic Analysis and Publication Prior Analysis

Objective: To improve the performance of citation recommendation, by using innovative citation recommendation methods. Methodology: By utilizing changes in the topical content of publications over time, citation graph, plus proximity-based citation contexts, I calculate the publication topical importance (prior) over time, and I use the topic prior, publication prior from author relation and Wikipedia along with the language model prioritize a number of scientific publication given their published time and user information need, i.e., a piece of working context or a publication abstract. Preliminary Experimentation and Result: In the preliminary experience, I employed the classical language model (without priors) and the language model with PageRank priors as the baseline. Preliminary experiment shows publication topic prior is important for citation recommendation given a textual information need.