Query-Driven Mining of Citation Networks for Patent Citation Retrieval and Recommendation
Parvaz Mahdabi, Fábio Crestani · 2014
Prior art search or recommending citations for a patent application is a challenging task. Many approaches have been proposed and shown to be useful for prior art search. However, most of these methods do not consider the network structure for integrating and diffusion of different kinds of information present among tied patents in the citation network. In this paper, we propose a method based on a time-aware random walk on a weighted network of patent citations, the weights of which are characterized by contextual similarity relations between two nodes on the network. The goal of the random walker is to find influential documents in the citation network of a query patent, which can serve as candidates for drawing query terms and bigrams for query refinement. The experimental results on CLEF-IP datasets (CLEF-IP 2010 and CLEF-IP 2011) show the effectiveness of encoding contextual similarities (common classification codes, common inventor, and common applicant) between nodes in the citation network. Our proposed approach can achieve significantly better results in terms of recall and Mean Average Precision rates compared to strong baselines of prior art search.