Detecting Symptoms using Context-based Twitter Embeddings during COVID-19

Roshan Santosh, Sharath Chandra Guntuku, H. Schwartz, johannes Eichstaedt, Lyle Ungar

Keywords: symptoms, twitter, unsupervised graph-based lexical expansion, social media

TL;DR: We present an unsupervised graph-based method for finding context-specific words and texts (e.g. symptoms) in large imbalanced corpora (e.g. tweets with #COVID-19), useful for syndromic surveillance of diseases whose pathology is evolving.

Abstract: In this paper, we present an unsupervised graph-based approach for the detection of symptoms of COVID-19, the pathology of which seems to be evolving. More generally, the method can be applied to finding context-specific words and texts (e.g. symptom mentions) in large imbalanced corpora (e.g. all tweets mentioning #COVID-19). Given the novelty of COVID-19, we also test the proposed approach generalizes to the problem of detecting Adverse Drug Reaction (ADR). We find that the approach applied on Twitter data can detect symptom mentions much prior to their being reported by the Centers for Disease Control (CDC).