**PaperWeaver: Enriching Topical Paper Alerts by Contextualizing Recommended Papers with User-collected Papers** **Authors:** Yoonjoo Lee, Hyeonsu B. Kang, Matt Latzke, Juho Kim, Jonathan Bragg, Joseph Chee Chang, and Pao Siangliulue **Abstract:** With the rapid growth of scholarly archives, researchers often subscribe to "paper alert" systems that provide recommendations of recently published papers similar to previously collected papers. However, these systems often only present paper titles and abstracts, making it difficult for researchers to understand the nuanced connections between recommended papers and their own research context. To address this issue, PaperWeaver is introduced, an enriched paper alerts system that provides contextualized text descriptions of recommended papers based on user-collected papers. PaperWeaver employs a computational method using Large Language Models (LLMs) to infer users' research interests from their collected papers, extract context-specific aspects of papers, and compare recommended and collected papers on these aspects. A user study (N=15) showed that participants using PaperWeaver were able to better understand the relevance of recommended papers and triage them more confidently compared to a baseline that presented related work sections from recommended papers. **Key Contributions:** - Qualitative findings from a formative study identified user challenges in making sense of recommended papers and the need for contextualized summaries. - PaperWeaver, a tool that provides additional contextualized descriptions for recommended papers tailored to user-collected papers. - Findings from a user study (N=15) demonstrated how using PaperWeaver facilitates sense-making of paper recommendations and aids in uncovering useful relationships between recommended and collected papers. **Design Goals:** 1. Describe details about recommended papers in a way that helps users understand their relevance to the user's research context. 2. Help users make connections between recommended and collected papers by comparing and contrasting them. 3. Reveal new aspects of previously collected papers to keep their understanding up-to-date and remind users of unread collected papers that have become more relevant. **Methods:** - **LLM-based Pipeline:** PaperWeaver generates three types of descriptions: contextualized aspect-based paper summaries, paper-paper descriptions based on citations, and paper-paper descriptions via generated pseudo-citations. - **Paper Alert Interface:** Users can explore different descriptions in three tabs: Related to Paper, Problem, method, and findings, and Abstract. **User Study:** - **Research Questions:** How do relevance descriptions augmented by user-collected papers' information help users understand the relevance of recommended papers, triage them, understand relationships between collected and recommended papers, and uncover new aspects of collected papers? - **Participants:** 15 researchers (Ph.D./MS students in CS domain). - **Design:** Within-subjects laboratory study. - **Results:** Participants using PaperWeaver were able to better understand the relevance of recommended papers and triage them more confidently**PaperWeaver: Enriching Topical Paper Alerts by Contextualizing Recommended Papers with User-collected Papers** **Authors:** Yoonjoo Lee, Hyeonsu B. Kang, Matt Latzke, Juho Kim, Jonathan Bragg, Joseph Chee Chang, and Pao Siangliulue **Abstract:** With the rapid growth of scholarly archives, researchers often subscribe to "paper alert" systems that provide recommendations of recently published papers similar to previously collected papers. However, these systems often only present paper titles and abstracts, making it difficult for researchers to understand the nuanced connections between recommended papers and their own research context. To address this issue, PaperWeaver is introduced, an enriched paper alerts system that provides contextualized text descriptions of recommended papers based on user-collected papers. PaperWeaver employs a computational method using Large Language Models (LLMs) to infer users' research interests from their collected papers, extract context-specific aspects of papers, and compare recommended and collected papers on these aspects. A user study (N=15) showed that participants using PaperWeaver were able to better understand the relevance of recommended papers and triage them more confidently compared to a baseline that presented related work sections from recommended papers. **Key Contributions:** - Qualitative findings from a formative study identified user challenges in making sense of recommended papers and the need for contextualized summaries. - PaperWeaver, a tool that provides additional contextualized descriptions for recommended papers tailored to user-collected papers. - Findings from a user study (N=15) demonstrated how using PaperWeaver facilitates sense-making of paper recommendations and aids in uncovering useful relationships between recommended and collected papers. **Design Goals:** 1. Describe details about recommended papers in a way that helps users understand their relevance to the user's research context. 2. Help users make connections between recommended and collected papers by comparing and contrasting them. 3. Reveal new aspects of previously collected papers to keep their understanding up-to-date and remind users of unread collected papers that have become more relevant. **Methods:** - **LLM-based Pipeline:** PaperWeaver generates three types of descriptions: contextualized aspect-based paper summaries, paper-paper descriptions based on citations, and paper-paper descriptions via generated pseudo-citations. - **Paper Alert Interface:** Users can explore different descriptions in three tabs: Related to Paper, Problem, method, and findings, and Abstract. **User Study:** - **Research Questions:** How do relevance descriptions augmented by user-collected papers' information help users understand the relevance of recommended papers, triage them, understand relationships between collected and recommended papers, and uncover new aspects of collected papers? - **Participants:** 15 researchers (Ph.D./MS students in CS domain). - **Design:** Within-subjects laboratory study. - **Results:** Participants using PaperWeaver were able to better understand the relevance of recommended papers and triage them more confidently