MUMBAI, India, Feb. 13 -- Intellectual Property India has published a patent application (202641008109 A) filed by Karpagam Academy Of Higher Education; Karpagam Institute Of Technology; Dr. V. Vadivu; and Vanathi Devi. M, Coimbatore, Tamil Nadu, on Jan. 28, for 'decepguard - detecting deceptive news using nlp.'
Inventor(s) include Dr. V. Vadivu; and Vanathi Devi. M.
The application for the patent was published on Feb. 13, under issue no. 07/2026.
According to the abstract released by the Intellectual Property India: "The rapid growth of social media has transformed the global information landscape, enabling real-time dissemination of news to millions of users. However, this convenience has also intensified the spread of fake news-misleading or fabricated content designed to manipulate public perception. The inability of traditional fact-checking methods to keep pace with high-volume online information has created an urgent need for automated, intelligent systems capable of detecting misinformation with high accuracy. To address this challenge, this article proposes WELFake, a two-phase benchmark model that integrates linguistic feature engineering with deep learning techniques to effectively classify news as real or fake. In the first phase, the proposed system preprocesses textual content and extracts a diverse set of linguistic features, including lexical, syntactic, and semantic attributes. These handcrafted features capture writing style, sentence structure, and word usage patterns that commonly distinguish fabricated news from authentic information. This phase serves as a robust baseline for evaluating textual integrity and provides interpretable insights into linguistic behaviors associated with deceptive content. The second phase incorporates a Convolutional Neural Network (CNN) architecture combined with word embedding techniques to learn deep semantic representations of the news text. By fusing linguistic features with CNN-generated semantic vectors, the model benefits from both surface-level linguistic cues and deep contextual understanding. This hybrid mechanism enhances the system's ability to detect nuanced patterns in deceptive news articles that may be overlooked by single-method approaches. A key contribution of this work is the development of the WELFake dataset, a novel composite dataset formed by merging multiple existing fake-news corpora. This consolidated dataset minimizes domain bias and ensures a more balanced and diverse training environment. Experimental evaluation demonstrates that the WELFake model achieves an accuracy of approximately 93%, outperforming several traditional baselines. The proposed approach offers a scalable and reliable solution for misinformation detection, contributing significantly to the integrity and trustworthiness of digital information ecosystems."
Disclaimer: Curated by HT Syndication.
