A survey on multi-lingual offensive language detection

Abstract

The prevalence of offensive content on online communication and social media platforms is growing more and more common, which makes its detection difficult, especially in multilingual settings. The term “Offensive Language” encompasses a wide range of expressions, including various forms of hate speech and aggressive content. Therefore, exploring multilingual offensive content, that goes beyond a single language, focus and represents more linguistic diversities and cultural factors. By exploring multilingual offensive content, we can broaden our understanding and effectively combat the widespread global impact of offensive language. This survey examines the existing state of multilingual offensive language detection, including a comprehensive analysis on previous multilingual approaches, and existing datasets, as well as provides resources in the field. We also explore the related community challenges on this task, which include technical, cultural, and linguistic ones, as well as their limitations. Furthermore, in this survey we propose several potential future directions toward more efficient solutions for multilingual offensive language detection, enabling safer digital communication environment worldwide.

Multilingual Hate Speech Detection: A Semi-Supervised Generative Adversarial Approach

“…In order to give a detailed overview of the concept, we aim to illustrate the definition of hate speech on social media. Hate speech is a specific subset of offensive language that directly targets individuals or groups based on specific features with the intent to discriminate or incite harm [9]. Generally, hate speech is defined as a conscious and intended public expression aimed at criticizing a specific group of people, whether based on race, religion, ethnicity, nationality, gender, sexual identity, or orientation.…”

“…Overall, we provide a definition of this phenomenon in a survey paper in which we established extensive study, and we define hate speech as any content that targets individuals or groups based on several factors such as race, ethnicity, religion, sexual orientation, gender, or other identifiable characteristics. This concept often reflects the policies and guidelines set by multiple social media platforms, which are influenced by legal frameworks and societal standards [9]. In addition, the spread of this content exceeds linguistic borders and encompasses more languages over time.…”

“…Among the most common challenges, is the cultural backgrounds, which affect the interpretation of this content, that impact its perception across various regions and populations, even within the same language. This complexity is made by the various dialects within languages like Arabic [9]. Moreover, users are becoming more familiar with the automatic detection algorithms, and they have discovered many ways to censor their hateful content to prevent its detection.…”

“…Most of the existing machine learning solutions (monolingual and multilingual) have used supervised learning approaches [9,11], where transfer learning techniques, based on Pretrained Language Models (PLMs), have proven to give outstanding results in multilingual hate speech detection. In fact, transformer-based architectures, such as BERT [14], have been demonstrated to achieve state-of-the-art performance in a variety of hate speech detection tasks.…”