The field of hate speech detection is rapidly advancing, with a focus on leveraging large language models and multimodal approaches to improve detection accuracy in low-resource languages and settings. Recent work has explored the use of translation-based approaches and contrastive learning to jointly align audio and text representations, resulting in improved performance in detecting hate speech in languages such as Urdu, Spanish, and low-resource Indian languages.
Notable research includes the development of a multimodal zero-shot framework for deepfake hate speech detection in low-resource languages, as well as the creation of large-scale datasets for hate speech detection in languages such as Portuguese. Additionally, the use of large language models has been evaluated for toxic language detection in low-resource languages, including Serbian, Croatian, and Bosnian.
In the field of online discourse analysis and moderation, researchers are developing innovative methods and tools to detect and mitigate extremist language, hate speech, and biased media coverage. This includes the creation of AI-based platforms for monitoring and fostering democratic discourse, as well as detecting neo-fascist rhetoric and other forms of hate speech. Notable research includes the development of a neo-fascist coding scheme for digital discourse in the USA societal context, as well as the creation of a national-scale infrastructure for monitoring political and media discourse across platforms in near real-time.
The field of multimodal analysis and detection is also rapidly evolving, with a focus on developing innovative methods to analyze and interpret complex data from various sources. Recent studies have explored the capabilities of large language models and vision-language models in detecting deception, image splicing, and deepfakes, as well as analyzing emotions and sentiment in text and images.
Furthermore, the field of natural language processing is witnessing significant advancements in multilingual language modeling and speech processing. Researchers are actively working on developing and evaluating large language models for low-resource languages, highlighting the need for more investment in these areas to address the performance gap between high-resource and low-resource languages.
Overall, these advancements demonstrate significant progress in the development of more effective and robust methods for hate speech detection, online discourse analysis, and multimodal analysis. As these fields continue to evolve, it is likely that we will see even more innovative solutions to these complex problems.
