Phonemer at WNUT-2020 Task 2: Sequence Classification Using COVID Twitter BERT and Bagging Ensemble Technique based on Plurality Voting

TL;DR

This paper develops an ensemble approach using COVID Twitter BERT and plurality voting to classify informative COVID-19 tweets, achieving high accuracy and ranking sixth in the WNUT-2020 shared task.

Contribution

It introduces a novel ensemble method combining multiple deep learning models with COVID Twitter BERT for tweet classification.

Findings

Achieved an F1-score of 0.9037

Outperformed individual models in ensemble

Ranked sixth in the shared task

Abstract

This paper presents the approach that we employed to tackle the EMNLP WNUT-2020 Shared Task 2 : Identification of informative COVID-19 English Tweets. The task is to develop a system that automatically identifies whether an English Tweet related to the novel coronavirus (COVID-19) is informative or not. We solve the task in three stages. The first stage involves pre-processing the dataset by filtering only relevant information. This is followed by experimenting with multiple deep learning models like CNNs, RNNs and Transformer based models. In the last stage, we propose an ensemble of the best model trained on different subsets of the provided dataset. Our final approach achieved an F1-score of 0.9037 and we were ranked sixth overall with F1-score as the evaluation criteria.