Extraction of Relevant Data from Social Media Based on Termino-Ontological Resources: Application to Meningitis Surveillance via Twitter

Abstract

In this paper, we present a process for collecting and filtering relevant data for epidemiological surveillance of meningitis. We focus on the African meningitis belt stretching from Senegal to Ethiopia. This study aims to fill the data gap for the early detection of epidemics based on the analysis of social media. Our approach is based on previous work that showed that social media analysis contributes significantly to the surveillance of epidemics. It uses IDOMEN (Infectious Disease Ontology for MENingitis) a meningitis domain ontology and a SKOS resource meningVocab (meningitis vocabulary). IDOMEN is an extension of the Infectious Disease Ontology (IDO). The SKOS resource meningVocab is built from a corpus of meningitis tweets from social media. We align the IDOMEN ontology and the SKOS resource meningVocab for collection and filtering tweets containing data relevant to meningitis in a perspective of epidemiological surveillance. Tweets are collected via the Twitter API on the basis of a list of terms related to meningitis. They are then annotated using these two resources and filtered using the rules of the domain (for example, the rules characterizing situations suggestive of bacterial meningitis: fever AND purpura AND headache).