Intellectual Property India Publishes Patent Application for A Dual-Path Convolutional Network With Hybrid Word Embeddings For Toxicity Detection

MUMBAI, India, Aug. 24 -- Intellectual Property India has published a patent application (202621081467 A) filed by Symbiosis International Deemed University on July 01, 2026, for A Dual-Path Convolutional Network With Hybrid Word Embeddings For Toxicity Detection.

Inventors include Dr. Aniket K. Shahade; Dr. Priyanka V. Deshmukh; Dr. Ketan Kotecha; and Dr. Shruti Patil.

The application for the patent was published on August 21, 2026, under issue no. 34/2026.

Abstract: ABSTRACT A DUAL-PATH CONVOLUTIONAL NETWORK WITH HYBRID WORD EMBEDDINGS FOR TOXICITY DETECTION The present invention discloses a dual-path CNN framework (100) for toxicity detection employing a hybrid embedding module (120) with static GloVe (122) and trainable (124) channels, a multichannel CNN module (130) with short-range (132, kernel=3) and long-range (134, kernel=5) convolutional pathways, a feature fusion module (140), and a classification module (150) with weighted binary cross-entropy loss (152) for class imbalance. The hybrid design captures both general semantics and platform-specific toxic patterns. The framework achieves 93.4 percent accuracy and 0.97 AUC-ROC with 380 comments per second throughput for real-time content moderation. [

Disclaimer: Curated by HT Syndication.