MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202611053055 A) filed by Sharda University on April 25, 2026, for Detection Of Targeted Speech Through Hierarchical Lexical Fusion And Counterfactual Analysis.
Inventors include Rawat, Paritosh; Chauhan, Priyanshu; Tyagi, Priyanshu; Kumar, Kapil; and Verma, Manju.
The application for the patent was published on July 10, 2026, under issue no. 28/2026.
Abstract: The present disclosure provides a system (100) and a method (300) for detection of targeted speech in an input text (102). The system (100) includes a processor (104), a memory (106) holding a pre-trained language model (106-2) and a hierarchical lexicon (106-4), an encoder (108) producing a contextual vector, a lexical scanner (110) producing a category vector with a weighted element per category corresponding to a count of matching sub-categories, a concatenator (112) producing a fused vector, a classifier (114) producing a class label, a counterfactual unit (116) interrogating contribution of category-bearing terms, and an explanation unit (118) producing an output (120). Unlike conventional approaches that treat category-bearing vocabulary as a flat undifferentiated signal, the present approach delivers fine-grained, causally interrogated, human-readable detection across multilingual content for content moderation, regulatory compliance, and evidentiary workflows.
Disclaimer: Curated by HT Syndication.
