A Word Graph-based Method for Disease Topic Identification in Biomedical Literature.

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Accurate identification of disease topics in biomedical literature is crucial for precise information retrieval.
• Traditional neural network approaches and disease mention frequency have limitations in disease topic identification.

Purpose of the Study:

• To propose a novel word graph-based method for enhanced disease topic identification in biomedical literature.
• To evaluate the performance of the proposed method against existing approaches.

Main Methods:

• Constructing word graphs from biomedical literature titles and abstracts.
• Extracting graph features for classification using logistic regression or random forest.
• Comparing the word graph method with disease mention frequency and hierarchical attention networks.

Main Results:

• Word graph features significantly outperformed disease mention frequency.
• The proposed method achieved superior performance compared to hierarchical attention networks.
• Demonstrated practical application in identifying disease topics.

Conclusions:

• The word graph-based method offers a more effective approach for disease topic identification in biomedical literature.
• This method enhances the precision of biomedical literature search and information retrieval.