ORCID

Abstract

Many new developments to detect and mitigate toxicity are currently being evaluated. Weare particularly interested in the correlation between toxicity and the emotions expressed inonline posts. While toxicity may be disguisedby amending the wording of posts, emotionswill not. Therefore, we describe here an ensemble method to identify toxicity and classify the emotions expressed on a corpus ofannotated posts published by Task 5 of SemEval 2021—our analysis shows that the majority of such posts express anger, sadness andfear. Our method to identify toxicity combinesa lexicon-based approach, which on its ownachieves an F1 score of 61.07%, with a supervised learning approach, which on its ownachieves an F1 score of 60%. When both methods are combined, the ensemble achieves an F1score of 66.37%.

Publication Date

2021-01-01

Event

The 15th International Workshop on Semantic Evaluation (SemEval-2021)

Publication Title

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021). Association for Computational Linguistics

Acceptance Date

2021-03-29

Deposit Date

2024-06-04

Embargo Period

2021-07-28

First Page

860

Last Page

864

Share

COinS