BioCaster

BioCaster detects early infectious disease outbreaks by automated text mining of global news media using ontology-based methodologies and neural network language models.

Key Features:

• Automated text mining: Processes open-source global news media to extract disease-related reports and events.
• Ontology-based methodologies: Applies ontology-based methods to represent and analyze disease concepts and relationships in text.
• Neural network language models: Integrates state-of-the-art neural language models for contextual and semantic understanding of news content.
• Neural Machine Translation: Supports translation across 10 languages to enable multilingual source processing.
• Neural Classification: Employs neural network classifiers to identify and categorize outbreak reports.
• Event semantics and risk evaluation: Incorporates event semantics to support assessment and evaluation of outbreak risk at multiple granularities.

Scientific Applications:

• Early outbreak detection: Identifies potential infectious disease threats from news reports to provide early warnings.
• Multilingual surveillance: Enables monitoring of disease-related information across multiple languages via neural machine translation.
• Event-level risk assessment: Facilitates assessment of outbreak risk using extracted event semantics at different levels of granularity.
• Evidence extraction for public health response: Extracts and classifies relevant event information to inform timely epidemiological and public health actions.

Methodology:

Automated text mining of global news media using ontology-based processing, neural machine translation (10 languages), neural network language models for contextual and semantic extraction, and neural classification of detected outbreak reports; extraction of event semantics for risk evaluation.