Towards role-based filtering of disease outbreak reports

Area of Science:

• Computational linguistics
• Bioinformatics
• Public Health Informatics

Background:

• Disease outbreak reports contain crucial information for public health protection.
• Effective text mining systems are essential for analyzing these reports.
• Named entities (NEs) represent key concepts within these reports.

Purpose of the Study:

• To explore the role of NEs in classifying disease outbreak reports.
• To investigate the impact of NE roles and semantic categories on text classification performance.
• To enhance the accuracy of text mining systems for public health.

Main Methods:

• Formal ontological methodology for conceptual analysis and classification of NEs.
• Extraction, combination, and feature engineering of NE roles.
• Integration of NE roles with semantic categories of disease-related nouns and verbs.
• Implementation of naïve Bayes and Support Vector Machine (SVM) algorithms for text classification.

Main Results:

• NE roles, when combined with NEs, significantly improve text classification performance.
• The combination of NE roles with semantic categories of noun and verb features substantially boosts classification accuracy.
• Both improvements were statistically significant compared to a baseline 'raw text' representation.

Conclusions:

• NE roles are valuable features for enhancing disease outbreak report classification.
• Integrating NE roles with semantic features offers a powerful approach for improving text mining in public health.
• The findings provide a foundation for developing more sophisticated and accurate public health surveillance systems.