NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets

Anders Giovanni Møller; Rob van der Goot; Barbara Plank

NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets

Anders Giovanni Møller, Rob van der Goot, Barbara Plank

NLPnorth

Research output: Conference Article in Proceeding or Book/Report chapter › Article in proceedings › Research › peer-review

Abstract

With the COVID-19 pandemic raging world-wide since the beginning of the 2020 decade,the need for monitoring systems to track relevant information on social media is vitally important. This paper describes our submission to the WNUT-2020 Task 2: Identification of informative COVID-19 English Tweets. We investigate the effectiveness for a variety of classification models, and found that domain-specific pre-trained BERT models lead to the best performance. On top of this, we attempt a variety of ensembling strategies, but these at-tempts did not lead to further improvements.Our final best model, the standalone CT-BERT model, proved to be highly competitive, leading to a shared first place in the shared task.Our results emphasize the importance of do-main and task-related pre-training.

Original language	English
Title of host publication	Proceedings of the Sixth Workshop on Noisy User-generated Text (W-NUT 2020)
Publisher	Association for Computational Linguistics
Publication date	Nov 2020
Pages	331-336
Publication status	Published - Nov 2020

Keywords

COVID-19 Pandemic
Social Media Monitoring
Informative Tweets
Pre-trained BERT Models
Domain-Specific Classification

Access to Document

2020.wnut-1.44Accepted author manuscript, 163 KB

https://www.aclweb.org/anthology/2020.wnut-1.44/

Multi-Task Sequence Labeling Under Adverse Conditions
Plank, B. (PI) & van der Goot, R. (CoI)
Amazon.com, Inc.
01/04/2019 → 31/08/2020
Project: Other

Cite this

@inproceedings{a4b8c1efbd87463ba6a510faadbec283,

title = "NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets",

abstract = "With the COVID-19 pandemic raging world-wide since the beginning of the 2020 decade,the need for monitoring systems to track relevant information on social media is vitally important. This paper describes our submission to the WNUT-2020 Task 2: Identification of informative COVID-19 English Tweets. We investigate the effectiveness for a variety of classification models, and found that domain-specific pre-trained BERT models lead to the best performance. On top of this, we attempt a variety of ensembling strategies, but these at-tempts did not lead to further improvements.Our final best model, the standalone CT-BERT model, proved to be highly competitive, leading to a shared first place in the shared task.Our results emphasize the importance of do-main and task-related pre-training.",

keywords = "COVID-19 Pandemic, Social Media Monitoring, Informative Tweets, Pre-trained BERT Models, Domain-Specific Classification, COVID-19 Pandemic, Social Media Monitoring, Informative Tweets, Pre-trained BERT Models, Domain-Specific Classification",

author = "M{\o}ller, {Anders Giovanni} and {van der Goot}, Rob and Barbara Plank",

year = "2020",

month = nov,

language = "English",

pages = "331--336",

booktitle = "Proceedings of the Sixth Workshop on Noisy User-generated Text (W-NUT 2020)",

publisher = "Association for Computational Linguistics",

address = "United States",

}

NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets. / Møller, Anders Giovanni; van der Goot, Rob ; Plank, Barbara.
Proceedings of the Sixth Workshop on Noisy User-generated Text (W-NUT 2020). Association for Computational Linguistics, 2020. p. 331-336.

Research output: Conference Article in Proceeding or Book/Report chapter › Article in proceedings › Research › peer-review

TY - GEN

T1 - NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets

AU - Møller, Anders Giovanni

AU - van der Goot, Rob

AU - Plank, Barbara

PY - 2020/11

Y1 - 2020/11

N2 - With the COVID-19 pandemic raging world-wide since the beginning of the 2020 decade,the need for monitoring systems to track relevant information on social media is vitally important. This paper describes our submission to the WNUT-2020 Task 2: Identification of informative COVID-19 English Tweets. We investigate the effectiveness for a variety of classification models, and found that domain-specific pre-trained BERT models lead to the best performance. On top of this, we attempt a variety of ensembling strategies, but these at-tempts did not lead to further improvements.Our final best model, the standalone CT-BERT model, proved to be highly competitive, leading to a shared first place in the shared task.Our results emphasize the importance of do-main and task-related pre-training.

AB - With the COVID-19 pandemic raging world-wide since the beginning of the 2020 decade,the need for monitoring systems to track relevant information on social media is vitally important. This paper describes our submission to the WNUT-2020 Task 2: Identification of informative COVID-19 English Tweets. We investigate the effectiveness for a variety of classification models, and found that domain-specific pre-trained BERT models lead to the best performance. On top of this, we attempt a variety of ensembling strategies, but these at-tempts did not lead to further improvements.Our final best model, the standalone CT-BERT model, proved to be highly competitive, leading to a shared first place in the shared task.Our results emphasize the importance of do-main and task-related pre-training.

KW - COVID-19 Pandemic

KW - Social Media Monitoring

KW - Informative Tweets

KW - Pre-trained BERT Models

KW - Domain-Specific Classification

KW - COVID-19 Pandemic

KW - Social Media Monitoring

KW - Informative Tweets

KW - Pre-trained BERT Models

KW - Domain-Specific Classification

M3 - Article in proceedings

SP - 331

EP - 336

BT - Proceedings of the Sixth Workshop on Noisy User-generated Text (W-NUT 2020)

PB - Association for Computational Linguistics

ER -

NLP North at WNUT-2020 Task 2: Pre-training versus Ensembling for Detection of Informative COVID-19 English Tweets

Abstract

Keywords

Access to Document

Fingerprint

Projects

Multi-Task Sequence Labeling Under Adverse Conditions

Cite this