Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble

Georgios Tziafas; Konstantinos Kogkalidis; Tommaso Caselli

doi:10.18653/v1/2021.nlp4if-1.18

Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble

Georgios Tziafas, Konstantinos Kogkalidis, Tommaso Caselli

ILS - Language, logic and information

University of Groningen

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

Abstract

This paper describes the TOKOFOU system, an ensemble model for misinformation detection tasks based on six different transformer-based pre-trained encoders, implemented in the context of the COVID-19 Infodemic Shared Task for English. We fine tune each model on each of the task’s questions and aggregate their prediction scores using a majority voting approach. TOKOFOU obtains an overall F1 score of 89.7%, ranking first.

Original language	English
Title of host publication	Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda
Publisher	Association for Computational Linguistics
Pages	119-124
DOIs	https://doi.org/10.18653/v1/2021.nlp4if-1.18
Publication status	Published - Jun 2021
Event	The Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda - Duration: 6 Jun 2021 → … http://www.netcopia.net/nlp4if/

Workshop

Workshop	The Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda
Period	6/06/21 → …
Internet address	http://www.netcopia.net/nlp4if/

Access to Document

10.18653/v1/2021.nlp4if-1.18Licence: CC BY

2021.nlp4if-1.18Final published version, 241 KBLicence: CC BY

Cite this

@inproceedings{afc6bc3a42c84b9da14a5999cb10a8d1,

title = "Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble",

abstract = "This paper describes the TOKOFOU system, an ensemble model for misinformation detection tasks based on six different transformer-based pre-trained encoders, implemented in the context of the COVID-19 Infodemic Shared Task for English. We fine tune each model on each of the task{\textquoteright}s questions and aggregate their prediction scores using a majority voting approach. TOKOFOU obtains an overall F1 score of 89.7%, ranking first.",

author = "Georgios Tziafas and Konstantinos Kogkalidis and Tommaso Caselli",

year = "2021",

month = jun,

doi = "10.18653/v1/2021.nlp4if-1.18",

language = "English",

pages = "119--124",

booktitle = "Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda",

publisher = "Association for Computational Linguistics",

address = "United States",

note = "The Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda ; Conference date: 06-06-2021",

url = "http://www.netcopia.net/nlp4if/",

}

Tziafas, G, Kogkalidis, K & Caselli, T 2021, Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble. in Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda. Association for Computational Linguistics, pp. 119-124, The Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda, 6/06/21. https://doi.org/10.18653/v1/2021.nlp4if-1.18

Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble. / Tziafas, Georgios; Kogkalidis, Konstantinos; Caselli, Tommaso.
Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda. Association for Computational Linguistics, 2021. p. 119-124.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble

AU - Tziafas, Georgios

AU - Kogkalidis, Konstantinos

AU - Caselli, Tommaso

PY - 2021/6

Y1 - 2021/6

N2 - This paper describes the TOKOFOU system, an ensemble model for misinformation detection tasks based on six different transformer-based pre-trained encoders, implemented in the context of the COVID-19 Infodemic Shared Task for English. We fine tune each model on each of the task’s questions and aggregate their prediction scores using a majority voting approach. TOKOFOU obtains an overall F1 score of 89.7%, ranking first.

AB - This paper describes the TOKOFOU system, an ensemble model for misinformation detection tasks based on six different transformer-based pre-trained encoders, implemented in the context of the COVID-19 Infodemic Shared Task for English. We fine tune each model on each of the task’s questions and aggregate their prediction scores using a majority voting approach. TOKOFOU obtains an overall F1 score of 89.7%, ranking first.

U2 - 10.18653/v1/2021.nlp4if-1.18

DO - 10.18653/v1/2021.nlp4if-1.18

M3 - Conference contribution

SP - 119

EP - 124

BT - Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda

PB - Association for Computational Linguistics

T2 - The Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda

Y2 - 6 June 2021

ER -

Fighting the COVID-19 Infodemic with a Holistic BERT Ensemble

Abstract

Workshop

Access to Document

Fingerprint

Cite this