QCRI at SemEval-2023 Task 3: News Genre, Framing and Persuasion Techniques Detection Using Multilingual Models

Proceedings of the The 17th International Workshop on Semantic Evaluation (SemEval-2023), pages 1237–1244 July 13-14, 2023 ©2023 Association for Computational Linguistics QCRI at SemEval-2023 Task 3: News Genre, Framing and Persuasion Techniques Detection using Multilingual Models Maram Hasanain1, Ahmed Oumar El-Shangiti1∗, Rabindra Nath Nandi2, Preslav Nakov3 and Firoj Alam1 1Qatar Computing Research Institute, HBKU, Qatar 2Hishab Singapore Pte. Ltd, Singapore, 3MBZUAI, UAE, maramhasanain@gmail.com, ahmedmohamedlemin@gmail.com, rabindro.rath@gmail.com, preslav.nakov@mbzuai.ac.ae, fialam@hbku.edu.qa Abstract Misinformation spreading in mainstream and social media has been misleading users in dif- ferent ways. Manual detection and verifica- tion efforts by journalists and fact-checkers can no longer cope with the great scale and quick spread of misleading information. This motivated research and industry efforts to de- velop systems for analyzing and verifying news spreading online. The SemEval-2023 Task 3 is an attempt to address several subtasks un- der this overarching problem, targeting writing techniques used in news articles to affect read- ers’ opinions. The task addressed three sub- tasks with six languages, in addition to three “surprise” test languages, resulting in 27 differ- ent test setups. This paper describes our par- ticipating system to this task. Our team is one of the 6 teams that successfully submitted runs for all setups. The official results show that our system is ranked among the top 3 systems for 10 out of the 27 setups. 1 Introduction Monitoring and analyzing news have become an im- portant process to understand how different topics (e.g., political) are reported in different news media and within and across countries. This has many important applications since the tone, framing, and factuality of news reporting can significantly affect public reactions toward social or political agen- das. A news piece can be manipulated on multiple aspects to sway readers’ perceptions and actions. Going beyond information factuality, other aspects include objectivity/genre, framing dimensions in- serted to steer the focus of the audience (Card et al., 2015), and propaganda techniques used to persuade readers towards a certain agenda (Barrón-Cedeno et al., 2019; Da San Martino et al., 2019a). News categorization is a well studied problem in the natural language processing field. Recently, re- search attention has focused on classifying news by factuality (Zhou and Zafarani, 2020; Nakov et al., 2021), or other related categorizations such as fake vs. satire news (Low et al., 2022; Golbeck et al., 2018). However, there have been efforts towards other classification dimensions. Card et al. (2015) developed a corpus of news articles annotated by 15 framing dimensions such as economy, capacity and resources, and fairness and equality, to support development of systems for news framing classifi- cation. Moreover, identifying propagandistic con- tent has gained a lot of attention over several do- mains including news (Barrón-Cedeno et al., 2019; Da San Martino et al., 2019a), social media (Alam et al., 2022) and multimodal content (Dimitrov et al., 2021a,b). The SemEval-2023 Task 3 shared task aims at motivating research in the aforementioned cate- gorization tasks, namely: detection and classifi- cation of the genre, framing, and the persuasion techniques in news …

Content truncated to 500 words