Clustering-based fusion for medical information retrieval

Area of Science:

• Information Science
• Medical Informatics

Background:

• The rapid growth of medical publications presents significant challenges for information retrieval.
• Existing data fusion methods prioritize effectiveness but neglect computational efficiency.

Purpose of the Study:

• To develop a method for selecting optimal subsets of retrieval systems for data fusion.
• To enhance both the effectiveness and efficiency of information retrieval in the medical domain.

Main Methods:

• A clustering-based approach using the Chameleon algorithm to group candidate retrieval systems.
• Sequential Forward Selection to choose representative systems from each cluster for fusion.

Main Results:

• The proposed method significantly outperforms baseline and state-of-the-art subset selection techniques.
• Over 10% improvement in Mean Average Precision and Recall-level Precision.
• Over 5% improvement in Precision at 10 documents and Mean Reciprocal Rank.

Conclusions:

• The clustering-based subset selection effectively optimizes data fusion for medical literature search.
• This approach balances retrieval performance with computational cost, offering a more efficient solution.