YNU-HPCC at SemEval-2021 Task 6: Combining ALBERT and Text-CNN for Persuasion Detection in Texts and Images

Proceedings of the 15th International Workshop on Semantic Evaluation (SemEval-2021), pages 1045–1050 Bangkok, Thailand (online), August 5–6, 2021. ©2021 Association for Computational Linguistics 1045 YNU-HPCC at SemEval-2021 Task 6: Combining ALBERT and Text-CNN for Persuasion Detection in Texts and Images Xingyu Zhu, Jin Wang and Xuejie Zhang School of Information Science and Engineering Yunnan University Kunming, China Contact: xyzhu@mail.ynu.edu.cn, {wangjin, xjzhang}@ynu.edu.cn Abstract In recent years, memes combining image and text have been widely used in social media, and memes are one of the most popular types of content used in online disinformation cam- paigns. In this paper, our study on the detec- tion of persuasion techniques in texts and im- ages in SemEval-2021 Task 6 is summarized. For propaganda technology detection in text, we propose a combination model of both AL- BERT and Text-CNN for text classification, as well as a BERT-based multi-task sequence la- beling model for propaganda technology cov- erage span detection. For the meme classifi- cation task involved in text understanding and visual feature extraction, we designed a paral- lel channel model divided into text and image channels. Our method1 achieved a good per- formance on subtasks 1 and 3. The micro F1- scores of 0.492, 0.091, and 0.446 achieved on the test sets of the three subtasks ranked 12th, 7th, and 11th, respectively, and all are higher than the baseline model. 1 Introduction The intentional shaping of information to promote a predetermined agenda is called propaganda. Propa- ganda uses psychological and rhetorical techniques to achieve its purpose. Propaganda techniques gen- erally include the use of logical fallacies and ap- peal to the emotions of the audience. In recent years, memes combining images and text have been widely used in social media, and the use of memes can easily and effectively attract a large number of users on social platforms. Memes are one of the most popular types of content used in online disinformation campaigns (Martino et al., 2020) , and memes applied in a disinformation campaign achieve their purpose of influencing users through 1The code of this paper is availabled at: https:// github.com/zxyqujing/SemEval-2021-task6 rhetorical and psychological techniques. Therefore, it is meaningful to research computational tech- niques for automatically detecting propaganda in particular content. The SemEval 2021 Task 6 (Dimitrov et al., 2021) consists of three subtasks: • Subtask 1 - Given only the “textual content” of a meme, identify which of the 20 techniques are used. The 20 techniques include appeal to authority, loaded language, and name calling or labeling. • Subtask 2: Given only the “textual content” of a meme, identify which of the 20 techniques are used along with the span(s) of the text covered by each technique. • Subtask 3: Given a meme, identify which of the 22 techniques are used for both the tex- tual and visual content of the meme. These 22 technologies include the 20 technologies in subtasks 1 and 2, and 2 technologies, i.e., transfer and appeal to (strong) emotions, are added. The detection of propaganda techniques in texts …

Content truncated to 500 words