Bullseye: Structured Passage Retrieval and Document Highlighting for Scholarly Search
This is an incremental improvement for researchers and scholars needing efficient access to specific information in academic papers.
The paper tackles the problem of improving scholarly search by developing the Bullseye system, which identifies relevant passages, restricts them to user-specified sections, and highlights them in PDFs, resulting in modest system effectiveness gains and substantial improvements in user effectiveness and effort.
We present the Bullseye system for scholarly search. Given a collection of research papers, Bullseye: 1) identifies relevant passages using any on-the-shelf algorithm; 2) automatically detects document structure and restricts retrieved passages to user-specifed sections; and 3) highlights those passages for each PDF document retrieved. We evaluate Bullseye with regard to three aspects: system effectiveness, user effectiveness, and user effort. In a system-blind evaluation, users were asked to compare passage retrieval using Bullseye vs. a baseline which ignores document structure, in regard to four types of graded assessments. Results show modest improvement in system effectiveness while both user effectiveness and user effort show substantial improvement. Users also report very strong demand for passage highlighting in scholarly search across both systems considered.