Biomedical ontology improves biomedical literature clustering performance: a comparison study

Area of Science:

• Biomedical Informatics
• Computer Science
• Information Retrieval

Background:

• Document clustering is vital for organizing and analyzing large volumes of biomedical literature.
• Existing clustering methods may not fully leverage the semantic richness of biomedical data.

Purpose of the Study:

• To evaluate the impact of a biomedical ontology on the performance of document clustering for biomedical literature.
• To compare the effectiveness and scalability of various clustering algorithms with and without ontology enhancement.

Main Methods:

• A comprehensive comparison of document clustering approaches including hierarchical clustering, Bisecting K-means, K-means, and Suffix Tree Clustering (STC).
• Evaluation of clustering quality and scalability using a biomedical literature dataset, with and without the integration of a biomedical ontology.

Main Results:

• Biomedical ontologies significantly enhance the clustering quality of biomedical documents.
• Bisecting K-means, K-means, and STC algorithms show performance improvements when utilizing the ontology.
• Hierarchical clustering algorithms, despite showing poorer overall quality, did not benefit from the ontology integration.

Conclusions:

• Integrating biomedical ontologies is a valuable strategy for improving biomedical literature clustering.
• The choice of clustering algorithm influences the degree to which ontology integration enhances performance.
• Ontology-enhanced clustering offers improved document retrieval and text mining capabilities in the biomedical domain.