Excited to share our #UIST2022 paper introducing FeedLens, a novel system that repurposes people’s research paper feeds on Semantic Scholar as lenses. Without any additional training, these lenses show relevance for all types of related entities, like authors and institutions. 🧵
Say you have 2 research paper feeds on the topics Interpretability and Mental Models. Using FeedLens, you can view the entities in any content list on Semantic Scholar (e.g., papers and authors in search results or author homepages) based on their relevance to your research feeds
FeedLens uses the simple yet powerful polymorphic lenses technique, which converts a lens based on one entity (e.g., research papers) into a polymorphic lens that can be applied to any type or collection of entities in a knowledge graph (e.g., authors, institutions, conferences)
Our user studies show that FeedLens can be seamlessly integrated into an existing production system. FeedLens features improve user engagement and reduce the cognitive demands of exploratory literature search, despite adding to the overall informational content on a page.
We believe our polymorphic lenses technique is easily applicable to not just Semantic Scholar and the scientific literature knowledge graph, but all recommender systems or applications that capture user preferences in some form.
Please read our paper for details on the polymorphic lenses technique and FeedLens design: arxiv.org/abs/2208.07531