Combining Citation Network Information and Text Similarity for Research Article Recommender Systems

IEEE Access

卷:10

Matthew M. Montemore¹  Julia A. Sterling² 

[1] Department of Chemical and Biomolecular Engineering, Tulane University, New Orleans, LA, USA;

[2] Department of Mathematics, Tulane University, New Orleans, LA, USA;

关键词: Scientific literature;    recommender systems;    search engines;    search methods;   

DOI  :  10.1109/ACCESS.2021.3137960

来源: DOAJ

【 摘 要 】

Researchers often need to gather a comprehensive set of papers relevant to a focused topic, but this is often difficult and time-consuming using existing search methods. For example, keyword searching suffers from difficulties with synonyms and multiple meanings. While some automated research-paper recommender systems exist, these typically depend on either a researcher’s entire library or just a single paper, resulting in either a quite broad or a quite narrow search. With these issues in mind, we built a new research-paper recommender system that utilizes both citation information and textual similarity of abstracts to provide a highly focused set of relevant results. The input to this system is a set of one or more related papers, and our system searches for papers that are closely related to the entire set. This framework helps researchers gather a set of papers that are closely related to a particular topic of interest, and allows control over which cross-section of the literature is located. We show the effectiveness of this recommender system by using it to recreate the references of review papers. We also show its utility as a general similarity metric between scientific articles by performing unsupervised clustering on sets of scientific articles. We release an implementation, ExCiteSearch (bitbucket.org/mmmontemore/excitesearch), to allow researchers to apply this framework to locate relevant scientific articles.

【 授权许可】

Unknown