BioCaster in 2021: automatic disease outbreaks detection from global news media

Area of Science:

• Public Health
• Computational Epidemiology
• Artificial Intelligence

Background:

• The BioCaster system was initially developed in 2008 for early disease detection using ontology-based text mining.
• After a six-year hiatus, the system has been relaunched to incorporate modern advancements.

Purpose of the Study:

• To systematically upgrade the BioCaster methodology with state-of-the-art neural network language models.
• To retain and enhance the system's capabilities in logical reasoning and automated early detection of infectious disease outbreaks.
• To present recent extensions and discuss future improvements for global public health benefit.

Main Methods:

• Implementation of neural machine translation across 10 languages.
• Development of neural classification models for disease outbreak reports.
• Creation of a new cloud-based visualization dashboard for data presentation.

Main Results:

• Successful relaunch of the BioCaster system with enhanced functionalities.
• Integration of advanced neural network models for improved text mining and disease detection.
• Introduction of a user-friendly visualization dashboard for public health professionals.

Conclusions:

• The upgraded BioCaster system offers enhanced capabilities for early infectious disease outbreak detection.
• Future work includes integrating risk assessment with event semantics and multi-granularity risk analysis.
• The enhanced system aims to significantly benefit the global public health community through improved surveillance.