Tackling Covid-19 Conspiracies on Twitter using BERT Ensembles, GPT-3 Augmentation, and Graph NNs

Korenčić, Damir; Grubišić, Ivan; De La Peña Sarracén, Gretel Liz; Toselli, Alejandro Hector; Chulvi, Berta; Rosso, Paolo (2023) Tackling Covid-19 Conspiracies on Twitter using BERT Ensembles, GPT-3 Augmentation, and Graph NNs. In: Hicks, Steven; García Seco De Herrera, Alba; Langguth, Johannes; Lommatzsch, Andreas; Andreadis, Stelios; Dao, Minh-Son; Martin, Pierre-Etienne; Hürriyetoğlu, Ali; Thambawita, Vajira; Nordmo, Tor-Arne; Vuillemot, Romain; Larson, Martha, (eds.) Working Notes Proceedings of the MediaEval 2022 Workshop. Aachen: CEUR Workshop Proceedings, CEUR Workshop Proceedings, pp. 243-247 .

PDF - Published Version - article Available under License Creative Commons Attribution. Download (999kB)

Abstract

We describe several approaches to text- and graph-based classification for detecting COVID-19 conspiracies on Twitter. We tackle the tasks of text classification with and without graph data, and classification of Twitter users based on user graph. To this end, we experiment with large transformer ensembles, GPT-3-based techniques, and a variety of graph neural networks. Our results demonstrate that transformer ensembling and GPT-3 text augmentation can improve performance and stability, and that richer graph data does not necessarily lead to improved performance.

Item Type:	Conference or workshop item published in conference proceedings (UNSPECIFIED)
Uncontrolled Keywords:	natural language processing; classification; graph-based classification; large language models
Subjects:	TECHNICAL SCIENCES > Computing > Artificial Intelligence
Divisions:	Division of Electronics
Depositing User:	Ivan Grubišić
Date Deposited:	15 Jan 2026 11:08
URI:	http://fulir.irb.hr/id/eprint/10848

Actions (login required)

View Item