Beyond Translation: Concept Mediation of STEM Terminology in Multilingual Search

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Navigating academic discovery systems is a critical barrier to equitable scholarly access for non-native English speakers. While general language proficiency is often cited as the primary obstacle, we argue that the principal impediment to retrieval is an architectural mismatch between international students’ conceptual frameworks and the authoritative English lexicons used by academic repositories. We report a qualitative study of 18 international students, 13 of whom were STEM majors at an R1 university, who completed think-aloud library search tasks and participated in semi-structured interviews. A simple translation interface is insufficient: the major bottleneck is concept-mediation friction, especially for STEM terminology, where a user's intent in one language must navigate translation tools and cultural shifts before reaching a system's database schema. From these findings, we outline a design direction: a concept-mediation layer over the library's existing authority vocabularies, using retrieval-augmented generation for matching and a Model Context Protocol server as the integration path into current discovery interfaces. This poster reports the study and its design rationale; a pilot is under construction.