MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202611060263 A) filed by Manipal University Jaipur on May 12, 2026, for A System And Method For Fine-Tuning Large Language Models For Hate Speech Detection In English And Hinglish.
Inventors include Mr. Bhawani Singh Rathore; and Dr. Sandeep Chaurasia.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: The present invention relates to a system for detecting hate speech in code-mixed text. The system comprises a collection of textual data from social media platforms, followed by preprocessing comprising normalization, transliteration handling, noise removal, and standardization of non-uniform spellings. The system employs a hybrid annotation approach involving automated labelling using rule- based or model-assisted techniques and subsequent expert validation to improve label accuracy. Transformer-based large language models, including BERT variants and generative models, are fine-tuned using a parameter-efficient technique comprising low-rank adaptation (LoRA). Fine-tuning is performed under resource-constrained conditions to reduce computational requirements. The system ensures uniform training parameters across models and evaluates performance using precision, recall, F1-score, and accuracy. The system provides a model for detecting implicit and context-dependent hate speech and handling domain shift through custom datasets.
Disclaimer: Curated by HT Syndication.
