AbstractExplorer: Leveraging Structure-Mapping Theory to Enhance Comparative Close Reading at Scale
summary
AbstractExplorer claims to let users form a corpus-level mental model of 1,000+ conference abstracts while preserving sentence detail by splitting abstracts into sentence roles and phrase-level roles and aligning analogous components across papers. It uses LLM-powered faceted comparative close reading, role highlighting, and structure-based ordering and alignment. An ablation eye-tracking study (N=24) and a summative study (N=16) show the three features work best together, with demos on CHI and a 200-paper WIST subset available.
AI-written summary · about
chapters (6)
- 0:00Introduction and motivation
- 1:06Close reading at scale concept
- 2:53Live demo of Abstract Explorer
- 5:00Evaluation studies and results
- 9:31Design implications and future work
- 10:13Summary and call for collaboration
related talks
10mCrowdsourcing Scholarly Discourse Annotations
29mFrom Keyword Chaos to Clarity: Conquering a Complex Systematic Review with Python
29mConquering PDFs- document understanding beyond plain text
51mHow to Write Papers So People Can Read Them
08mComparing Apples and Oranges: Human and Computer Clustered Affinity Diagrams Under the Microscope
37mBuilding a Scientific Taxonomy at Scale
21mCloseread: bringing Scrollytelling to Quarto
18mGraphScape: A Model for Automated Reasoning about Visualization Similarity and Sequencing