Institutional Repository of Vilnius University
Hybrid retrieval methods and an attention mechanism in medical retrieval-augmented generation systems /
Abstract
dc:descriptionThis thesis evaluates information retrieval methods for medical document collections in a retrievalaugmented generation (RAG) setting. Clinical guideline documents were preprocessed to extract text, tables and images into a unified textual format, and multiple retrieval strategies were implemented and compared. Experimental results showed that dense semantic retrieval alone provides limited accuracy (33%), while subunitbased scoring improves early ranks (42%) and hybrid semanticlexical retrieval achieves the strongest performance (88%). Native sparse attention reranking did not yield meaningful improvements, when trained with a small subset of questionanswer pairs (35%).
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Braun, Ugnius Byron,
Subjects
dc:subject × 1Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- eng
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277678336&prefLang=en_US