Automatically created concept graphs using descriptive keywords in the medical domain.

Area of Science:

• Information Science
• Computer Science
• Medical Informatics

Background:

• Navigational search is crucial for digital object collections.
• Manual creation of categorization systems is costly and time-consuming.
• Automated methods are needed for efficient information retrieval.

Purpose of the Study:

• To introduce the Semantic GrowBag algorithm for automatic concept graph creation.
• To evaluate the algorithm's effectiveness in the medical domain.
• To compare GrowBag graphs with existing thesauri.

Main Methods:

• The Semantic GrowBag algorithm utilizes descriptive keywords and higher-order co-occurrences.
• Over 2000 GrowBag graphs were generated using the Medline dataset.
• Algorithm performance was compared against a baseline and the MeSH thesaurus.

Main Results:

• The Semantic GrowBag approach increased relevant relationships by approximately 50% over a baseline.
• Identified relationships in GrowBag graphs largely align with MeSH thesaurus.
• The algorithm demonstrates applicability in the medical domain.

Conclusions:

• Semantic GrowBag enables fully automated concept graph generation.
• GrowBag graphs effectively support navigation in digital object collections.
• The algorithm can aid in maintaining and updating existing categorization systems.