Deriving Disinformation Insights from Geolocalized Twitter Callouts

TL;DR

This paper presents a two-stage method combining geospatial classification and language modeling to analyze disinformation in Twitter data across European languages, revealing geographic and linguistic differences.

Contribution

It introduces a novel language-independent geolocation method and an analytical approach using lexical features and embeddings for disinformation analysis.

Findings

Effective classification of European vs. non-European tweets

Identification of geographic and linguistic differences in disinformation

Dataset of 36 million disinformation-related tweets in three languages

Abstract

This paper demonstrates a two-stage method for deriving insights from social media data relating to disinformation by applying a combination of geospatial classification and embedding-based language modelling across multiple languages. In particular, the analysis in centered on Twitter and disinformation for three European languages: English, French and Spanish. Firstly, Twitter data is classified into European and non-European sets using BERT. Secondly, Word2vec is applied to the classified texts resulting in Eurocentric, non-Eurocentric and global representations of the data for the three target languages. This comparative analysis demonstrates not only the efficacy of the classification method but also highlights geographic, temporal and linguistic differences in the disinformation-related media. Thus, the contributions of the work are threefold: (i) a novel language-independent transformer-based geolocation method; (ii) an analytical approach that exploits lexical specificity and word embeddings to interrogate user-generated content; and (iii) a dataset of 36 million disinformation related tweets in English, French and Spanish.