Analyzing Persuasive Strategies in Meme Texts: A Fusion of Language Models with Paraphrase Enrichment

TL;DR

The paper tackles detecting lexicalized persuasion techniques in memes by framing it as a hierarchical multi-label problem and evaluating zero-shot performance in surprise languages. It proposes an ensemble of fine-tuned BERT, XLM-RoBERTa, and mBERT models trained on paraphrase-augmented data, with per-class thresholds and a translation-based zero-shot pipeline. Key findings show that paraphrase augmentation improves performance and that balancing the training data yields stronger hierarchical F1 than simply expanding unbalanced data, though cross-distribution paraphrases can introduce noise and harm zero-shot results. These insights offer practical guidance for cross-domain and cross-language persuasion technique detection and highlight avenues for improving data augmentation, balancing strategies, and multilingual modeling.

Abstract

This paper describes our approach to hierarchical multi-label detection of persuasion techniques in meme texts. Our model, developed as a part of the recent SemEval task, is based on fine-tuning individual language models (BERT, XLM-RoBERTa, and mBERT) and leveraging a mean-based ensemble model in addition to dataset augmentation through paraphrase generation from ChatGPT. The scope of the study encompasses enhancing model performance through innovative training techniques and data augmentation strategies. The problem addressed is the effective identification and classification of multiple persuasive techniques in meme texts, a task complicated by the diversity and complexity of such content. The objective of the paper is to improve detection accuracy by refining model training methods and examining the impact of balanced versus unbalanced training datasets. Novelty in the results and discussion lies in the finding that training with paraphrases enhances model performance, yet a balanced training set proves more advantageous than a larger unbalanced one. Additionally, the analysis reveals the potential pitfalls of indiscriminate incorporation of paraphrases from diverse distributions, which can introduce substantial noise. Results with the SemEval 2024 data confirm these insights, demonstrating improved model efficacy with the proposed methods.

Figures (8)

• Figure 1: A sample training instance. The text is labelled with three techniques, Loaded Language, Slogans and Name calling/Labelling.
• Figure 2: Hierarchy of the persuasion techniques to be used to label the texts. Figure slightly modified from semeval2024task4.
• Figure 3: Distribution of the data for each persuasion technique in the training set.
• Figure 4: Distribution of persuasion techniques per instance in the training set.
• Figure 5: Schematic overview of our classification pipeline for the detection of persuasion techniques in memes.