This study evaluates OpenAI’s ChatGPT, a large language model, for its efficacy in detecting hate speech in Portuguese tweets, comparing it with purpose-trained models. Despite incurring considerable computational costs, ChatGPT as a zero-shot classifier demonstrated commendable performance, even superior to or on par with state-of-the-art methods, with an F1-score of 73.0% on the ToLD-BR. In a cross-dataset evaluation on the HLPHSP dataset, it secured a superior F1-score of 73%. The choice of prompt significantly impacts the outcome, with a wider scope prompt balancing precision and recall metrics. ChatGPT, due to its interpretability and resilience against data distribution shifts, could be a preferred choice for tasks prioritizing these factors.
The recognition of toxic and hate speech on social media platforms is important due to the significant risks posed to users and the digital ecosystem. Current state-of-the-art models, such as BERTimbau, have set benchmarks for Portuguese text classification, yet challenges remain in accurately detecting toxic content. This paper investigates the effectiveness of fine-tuning a smaller, open-source decoder-only model, LLaMA 3.1 8B 4bit, for this task. We propose an iterative prompt evolution method to optimize the model’s performance. Our results demonstrate that fine-tuning significantly enhances the LLaMA model’s F1-score from 0.61 to 0.75, surpassing BERTimbau in precision and matching the performance of the GPT-4o mini. However, the approach depends on the quality of the language models used for prompt evolution, highlighting the need for further research to enhance robustness in this area.