Team LEGO at SemEval-2022 Task 4: Machine Learning Methods for PCL Detection

Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022), pages 369 - 373 July 14-15, 2022 ©2022 Association for Computational Linguistics Team LEGO at SemEval-2022 Task 4: Machine Learning Methods for PCL Detection Abhishek Kumar Singh GLA University, India sabhishek.kumar166@gmail.com Abstract In this paper, we present our submission to the SemEval 2022 - Task 4 on Patronizing and Condescending Language (PCL) detection. We approach this problem as a traditional text clas- sification problem with machine learning (ML) methods. We experiment and investigate the use of various ML algorithms for detecting PCL in news articles. Our best methodology achieves an F1- Score of 0.39 for subtask1 with a rank of 63 out of 80, and F1-score of 0.082 for subtask2 with a rank of 41 out of 48 on the blind dataset provided in the shared task. 1 Introduction The explosion of social media in recent years also enables increasing the number of patronizing and condescending language (PCL). Patronizing and condescending language depicts apparently kind or helpful behavior but betraying a feeling of su- periority on others. Previously, harmful behavior in language for example, hate speech (Fortuna and Nunes, 2018), offensive language (Razavi et al., 2010), fake news (Oshikawa et al., 2018), rumor propagation or misinformation (Zhou and Zafarani, 2020), and many others has been widely studied in NLP, PCL has been a neglected area of study until very recently. Identifying PCL is hard even for humans be- cause it is subjective and subtle. For instance, one might find condescending something which another person might consider an objective portrayal of a situation or some people might not see the harm in describing how those in a privileged position donate their remainings to those who need them. Also, we would expect a member of a so-called vulnerable community to feel more patronised than one person who does not belong to such group while reading how others refer to them. The goal of SemEval 2022-Task 4 is to design a system to detect whether or not the text contains any form of PCL and furthermore, identify which Class Nb of Samples PCL 9476 Non-PCL 993 Table 1: Dataset Statistics Task 1 PCL category expresses the condescension. The organizers provided two datasets, one annotated based on the intensity of PCL and other with the PCL categories. We approach this problem using various machine learning approaches using the lin- guistics features. The structure for the rest of the paper is as fol- lows. Section 2 describes a background about the dataset. Section 3 describes the experimental setup of our experiments. It involves pre-processing, fea- ture engineering, implementation details for all the respective ML models. Section 4 describes the results and discussion for both the subtask. And lastly, in Section 5, we concluded the paper and suggest ideas for future research. 2 Dataset The dataset (Perez-Almendros et al., 2020) pro- vided for this challenge was collected from the News on Web (NoW) corpus (Davies, 2013). For task1, we are provided with 10469 text paragraphs. Each …

Content truncated to 500 words