TwiTrace: A New Approach of Contact Tracing Based on Multilingual Tweets
Faiza Deghmani, Kamel Boukhalfa · 2023
By the end of 2019, the world has known a resurgence of epidemics by the sudden outbreak of the COVID-19 virus, that claimed millions of lives, and overwhelmed public health systems. Compared to other known viruses, this one is thought to be more contagious. An individual may be infectious without showing symptoms, so until he tests positive, he might infect many people who meet him. Therefore, to keep the numbers under control, governments apply various strategies such social distancing, remote work and investigative methods like contact tracing to identify potential infected people. Many researchers proposed effective digital contact tracing solutions; nevertheless, they rely on user's cooperation to install applications or to carry sensor devices. However, social networks present a suitable alternative to gather contact-tracing data since they are public and available. In this paper, we propose TwiTrace a new contact tracing approach based on multilingual tweets and graph modelling. TwiTrace creates a dataset for Covid-19 positive cases using Twitter API, and then identifies close contacts and risk places. Next, a graph is generated and queried using Neo4j database. Our approach indicates high accuracy and shows good results in terms of size compared to suspected cases and high-risk places extracted manually.