Deriving Disinformation Insights from Geolocalized Twitter Callouts
arXiv:2108.03067
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.
arXiv commentAccepted for presentation at KDD 2021 - Workshop On Deriving Insights From User-Generated Text
References in corpus (7)
- Scikit-learn: Machine Learning in Python
- Distributed Representations of Words and Phrases and their Compositionality
- Geotagging One Hundred Million Twitter Accounts with Total Variation Minimization
- emoji2vec: Learning Emoji Representations from their Description
- Beto, Bentz, Becas: The Surprising Cross-Lingual Effectiveness of BERT
- Are All Languages Created Equal in Multilingual BERT?
- How COVID-19 Is Changing Our Language : Detecting Semantic Shift in Twitter Word Embeddings