PaperWeaver: Enriching Topical Paper Alerts by Contextualizing Recommended Papers with User-collected Papers

With the rapid growth of scholarly archives, researchers subscribe to "paperalert" systems that periodically provide them with recommendations of recentlypublished papers that are similar to previously collected papers. However,researchers sometimes struggle to make sense of nuanced connections betweenrecommended papers and their own research context, as existing systems onlypresent paper titles and abstracts. To help researchers spot these connections,we present PaperWeaver, an enriched paper alerts system that providescontextualized text descriptions of recommended papers based on user-collectedpapers. PaperWeaver employs a computational method based on Large LanguageModels (LLMs) to infer users' research interests from their collected papers,extract context-specific aspects of papers, and compare recommended andcollected papers on these aspects. Our user study (N=15) showed thatparticipants using PaperWeaver were able to better understand the relevance ofrecommended papers and triage them more confidently when compared to a baselinethat presented the related work sections from recommended papers.