Document Retrieval for Precision Medicine Using a Deep Learning Ensemble Method.

Area of Science:

• Biomedical Informatics
• Information Retrieval
• Machine Learning

Background:

• The rapid growth of biomedical literature presents significant challenges for researchers seeking specific information.
• Effective information retrieval is crucial but complicated by the need for multi-faceted relevance evaluation.

Purpose of the Study:

• To develop a systematic and improved method for retrieving scientific literature tailored to individual patient needs.
• To enhance the accuracy and efficiency of biomedical information retrieval systems.

Main Methods:

• Implemented query expansion and query boosting for initial document retrieval and ranking.
• Utilized a text classification model and a relevance matching model for multi-dimensional document evaluation.
• Combined model outputs using logistic regression for a refined document re-ranking process.

Main Results:

• The proposed ensemble method significantly improved biomedical retrieval performance.
• Achieved state-of-the-art results on the Text Retrieval Conference 2019 Precision Medicine Track dataset.
• Demonstrated superior performance compared to existing deep learning-based retrieval methods.

Conclusions:

• A novel deep learning-based ensemble method was successfully developed for biomedical literature retrieval.
• Query expansion and boosting strategies proved effective in the initial retrieval stages.
• Text classification and relevance matching models enhanced semantic understanding and improved overall retrieval accuracy.