NLPIITR at SemEval-2021 Task 6: RoBERTa Model with Data Augmentation for Persuasion Techniques Detection

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1061–1067 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 1061 NLPIITR at SemEval-2021 Task 6: RoBERTa Model with Data Augmentation for Persuasion Techniques Detection Vansh Gupta Indian Institute of Technology, Delhi vansh.g0108@gmail.com Raksha Sharma Indian Institute of Technology, Roorkee raksha.sharma@cs.iitr.ac.in Abstract This paper describes and examines different systems to address Task 6 of SemEval-2021: Detection of Persuasion Techniques In Texts And Images, Subtask 1. The task aims to build a model for identifying rhetorical and psycho- logical techniques (such as causal oversimplifi- cation, name-calling, smear) in the textual con- tent of a meme which is often used in a dis- information campaign to influence the users. The paper provides an extensive comparison among various machine learning systems as a solution to the task. We elaborate on the pre-processing of the text data in favor of the task and present ways to overcome the class imbalance. The results show that fine-tuning a RoBERTa model gave the best results with an F1-Micro score of 0.51 on the development set. 1 Introduction The purpose of this task was to identify propaganda in electoral campaigns where memes are one of the most popular types of content used to carry out disinformation campaigns. They are most effec- tive on social media platforms since they can easily reach many users. SemEval-2021 task 6 (Dimitrov et al., 2021) refers to propaganda whenever infor- mation is purposefully shaped to foster a predeter- mined plan. Propaganda uses psychological and rhetorical techniques to achieve its objectives. Such techniques include the use of logical fallacies and appealing to the emotions of the audience. Logical fallacies are usually hard to spot since the argu- mentation, at first sight, might seem correct and objective. However, a careful analysis shows that the conclusion cannot be drawn from the premise without the misuse of logical rules. Another set of techniques uses emotional language to induce the audience to agree with the speaker only based on the emotional bond that is being created, provok- ing the suspension of any rational analysis of the argumentation. The Oxford dictionary defines a meme as ‘An image, video, piece of text, etc., typically humor- ous in nature, that is copied and spread rapidly by internet users, often with slight variations’. They generally consist of an image superimposed with text. The role of the image in a deceptive meme is either to reinforce/complement a technique in the text or to convey one or more persuasion tech- niques. A total of 3 subtasks were defined: For subtasks 1 and 2, only the meme’s textual content was given. The former was a multilabel classifi- cation problem while the latter was a multilabel sequence tagging problem, each with 20 labels of techniques. For subtask 3, both textual and visual content of the memes were given. The goal was to identify which of the techniques are used for a given text/image. In this paper, we describe a system to address the …

Content truncated to 500 words