Abstract Cyberbullying has experienced a surge in recent years, mainly due to the widespread adoption of social media platforms. This trend manifests in multiple ways, with hostile language being one of the most common. The latter underscores the urgent need for robust detection methods to address this issue effectively. To address this problem, we propose a novel pipeline to enhance hostile language detection in social media. Our approach consists of a combination of two ideas: First, we propose conducting a Domain Adaptation procedure to specialize the knowledge of a pre-trained BERT, making it more specialized in the domain of social media. For this adaptation, we modify the traditional random Masked Language Modeling technique and propose three novel strategies for selecting the subset of tokens to mask out cleverly. Second, we tailor an Adversarial Regularizer when fine-tuning the adapted BERT for specific hostile-language datasets. We evaluate the performance of our method for detecting hate speech, aggressiveness, offensiveness, and sexism. Our results show that the Domain Adaptation procedure significantly outperforms vanilla BERT, and the Adversarial Regularizer can lead to more robust fine-tuning, thereby enhancing performance. Moreover, we demonstrate that these methods can be used together to achieve an even more significant performance boost.

Emilio Cueva, Adrian Lopez Monroy, Fernando Sánchez-Vega, Thamar Solorio. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Multimodal machine learning is a cutting-edge field that explores ways to incorporate information from multiple sources into models.As more multimodal data becomes available, this field has become increasingly relevant.This work focuses on two key challenges in multimodal machine learning.The first is finding efficient ways to combine information from different data types.The second is that often, one modality (e.g., text) is stronger and more relevant, making it difficult to identify meaningful patterns in the weaker modality (e.g., image).Our approach focuses on more effectively exploiting the weaker modality while dynamically regularizing the loss function.First, we introduce a new two-stream model called Multimodal BERT-ViT, which features a novel intra-CLS token fusion.Second, we utilize a dynamic adjustment that maintains a balance between specialization and generalization during the training to avoid overfitting, which we devised.We add this dynamic adjustment to the Unsupervised Data Augmentation (UDA) framework.We evaluate the effectiveness of these proposals on the task of multi-label movie genre classification using the Moviescope and MM-IMDb datasets.The evaluation revealed that our proposal offers substantial benefits, while simultaneously enabling us to harness the weaker modality without compromising the information provided by the stronger.