Conference paper · 2026 · Submitted
Information retrieval over academic curricula: the case of the Lattes platform
Jonathas Castilho
SIICUSP 2026, University of São Paulo International Symposium of Undergraduate Research
Abstract
Search over academic curricula mixes names, research areas and free-text biographies, which behave differently under lexical ranking. We evaluate BM25 against language models with Dirichlet smoothing on the USP Aberta search subsystem, measuring nDCG@3 across classes of queries over real faculty data.
Keywordsinformation retrieval · BM25 · language models · nDCG · Lattes
How to cite
@inproceedings{castilho2026information,
author = {Jonathas Castilho},
title = {Information retrieval over academic curricula: the case of the Lattes platform},
year = {2026},
booktitle = {SIICUSP 2026, University of São Paulo International Symposium of Undergraduate Research},
note = {Submitted}
}