MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112106 A) filed by Vallurupalli Nageswara Rao Vignana Jyothi Institute Of Engineering And Technology on September 18, 2026, for An Artificial Intelligence-Based Multilingual Text Abuse Detection System Using Bert-Based Natural Language Processing For Social Media Safety.
Inventor includes Mrs. Naga Durga Saile. K.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: ABSTRACT [0019] A system is presented for identifying abusive comments written in transliterated Telugu on social media. Comments are collected from Instagram, cleaned of special characters and noise, and manually labelled into three categories: no abuse, hate speech, or combined hate speech and abuse. A multilingual BERT model is fine-tuned on this data together with several neural architectures including feed-forward networks, recurrent networks, long short-term memory units, gated recurrent units and transformer encoders. The individual predictions are then combined by a meta-level feed-forward network that produces the final classification. The complete pipeline is exposed through a simple Flask web interface that accepts a comment and returns its predicted category in real time. Experiments show that the ensemble reaches a training accuracy of ninety-three point four five percent, demonstrating effective detection of culturally specific abusive patterns that conventional English-centric tools miss.
Disclaimer: Curated by HT Syndication.
