Improving biomedical information retrieval by linear combinations of different query expansion techniques

Area of Science:

• Computer Science
• Information Science
• Biomedical Informatics

Background:

• Biomedical literature retrieval faces challenges due to increasing data complexity.
• Current information retrieval (IR) systems often yield low-quality results, failing to precisely meet user information needs.
• Users struggle to construct accurate queries for retrieving specific data from large document repositories.

Purpose of the Study:

• To enhance the performance of information retrieval systems in the biomedical domain.
• To refine IR searches by better representing user information needs through query expansion.
• To introduce a novel technique combining multiple query expansion methods.

Main Methods:

• Developed a system employing various query expansion techniques.
• Applied linear combinations to pairs of query expansion results.
• Tested the system's effectiveness in refining information retrieval searches.

Main Results:

• The combined query expansion approach significantly improved retrieval performance.
• The proposed system outperformed the baseline by 21.06% in retrieval accuracy (MAP).
• The method also showed improvement over a previous study, outperforming it by 7.12%.

Conclusions:

• Proposed and validated a novel approach using linear combinations of query expansion techniques.
• The method enhances the cognizability of user queries for search engines.
• Achieved higher-quality search results, addressing limitations of basic IR systems.