Monge is a prototype of a geographical disease monitor based on tweets. By retrieving tweets located in different Spanish cities, both in Spanish and Catalan and processing and analyzing the information with Human Language Technologies techniques and tools, it allows predicting possible epidemic outbreaks of different diseases of general interest (flu, asthma, etc.).
Tweets are filtered using three criteria: location, language, and bags of disease words that have been generated using WordReference synonyms and embeddings. This demo could be very useful because the Centers for Disease Control and Prevention takes 1-2 weeks from the time a patient is diagnosed until the data is available, whereas this prototype offers real-time monitoring.