Monday 08 September 2025
A team of researchers has made significant progress in developing a system that can detect hate speech on social media platforms with high accuracy. The study, published in a recent scientific paper, used a large dataset of over 1.2 million social media posts to fine-tune a type of artificial intelligence (AI) called transformers.
The researchers found that one particular transformer model, called ELECTRA, outperformed all other models tested in detecting hate speech. The model achieved an F1 score of 0.8980, indicating its ability to accurately identify both true and false positive instances.
Hate speech on social media is a widespread problem, with the potential to cause significant harm to individuals and communities. It can take many forms, including explicit language, coded language, and indirect expressions. The detection of hate speech is a complex task due to the subjective nature of what constitutes hate speech.
The researchers used a dataset called MetaHate, which was compiled from 36 different sources and contains over 1.2 million social media posts. The dataset includes both hate speech and non-hate speech posts, allowing the researchers to evaluate the performance of their model.
To fine-tune the ELECTRA model, the researchers used a process called transfer learning. This involves pre-training the model on a large dataset before applying it to the specific task of detecting hate speech. They then used a technique called fine-tuning to adjust the model’s parameters based on the MetaHate dataset.
The results of the study show that the ELECTRA model is highly effective in detecting hate speech, with an accuracy rate of 0.8946 and an F1 score of 0.8980. The model also performed well across different types of hate speech, including explicit language and coded language.
The researchers acknowledge that their system is not without its limitations. For example, the dataset used to fine-tune the model may contain biases, which could affect the performance of the model in real-world applications. Additionally, the detection of hate speech is a complex task that requires a deep understanding of the context in which the language is being used.
Despite these limitations, the study’s findings have significant implications for the development of systems that can detect and prevent hate speech on social media platforms. The researchers hope that their work will contribute to the creation of more effective tools for combating online harassment and promoting a safer and more respectful online environment.
Cite this article: “Highly Accurate Hate Speech Detection System Developed Using Artificial Intelligence”, The Science Archive, 2025.
Here Are The Keywords: Hate Speech, Social Media, Artificial Intelligence, Transformers, Electra Model, Metahate Dataset, Transfer Learning, Fine-Tuning, Online Harassment, Language Detection
