BioCaster: detecting public health rumors with a Web-based text mining system

Area of Science:

• Computational epidemiology
• Public health informatics
• Text mining

Background:

• The BioCaster ontology provides multilingual background knowledge for disease and location identification.
• It bridges layman terms with formal coding systems, aiding in epidemiological analysis.

Purpose of the Study:

• To develop and evaluate an ontology-based text mining system for detecting and tracking infectious disease outbreaks.
• To leverage linguistic signals from the web for real-time public health surveillance.

Main Methods:

• Continuous analysis of over 1700 RSS feeds for relevant documents.
• Application of topic classification, named entity recognition (NER), and event recognition.
• Utilizing the BioCaster ontology for disease/location detection and geocoding on a Google map.

Main Results:

• The system successfully classifies documents, identifies diseases and geographical locations, and plots outbreaks.
• Higher-order event analysis enables precise warning signals.
• Evaluation on a gold standard corpus demonstrates effectiveness in topic and entity identification.

Conclusions:

• BioCaster is an effective system for detecting and tracking infectious disease outbreaks using web-based linguistic signals.
• The ontology-based approach enhances the accuracy and scope of public health surveillance.
• The system provides timely alerts for registered users, aiding in outbreak response.