Taygete at SemEval-2022 Task 4: RoBERTa based models for detecting Patronising and Condescending Language

TL;DR

This paper presents RoBERTa-based models combined with LSTM and CNN layers to detect patronising and condescending language in news articles, achieving top-15 rankings in SemEval 2022 Task-4.

Contribution

It introduces a novel combination of RoBERTa with LSTM and CNN for detecting condescending language, demonstrating competitive performance.

Findings

Achieved 15th place with F1-score 0.5924 in subtask-A

Secured 12th place with macro-F1 0.3763 in subtask-B

Showed effectiveness of RoBERTa-based models for nuanced language detection

Abstract

This work describes the development of different models to detect patronising and condescending language within extracts of news articles as part of the SemEval 2022 competition (Task-4). This work explores different models based on the pre-trained RoBERTa language model coupled with LSTM and CNN layers. The best models achieved 15$^{th}$ rank with an F1-score of 0.5924 for subtask-A and 12$^{th}$ in subtask-B with a macro-F1 score of 0.3763.