MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641047772 A) filed by Francis Xavier Engineering College, Tirunelveli, Tamil Nadu, on April 15, for 'real-time fake news detection platform with automated web scraping and ml-based prediction.'
Inventor(s) include S. Arunadevi; M. Harrish; M. Kaliraja; and K. Esakki Muthu.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "A real-time fake news detection platform is disclosed, comprising an automated web scraping engine, a three-tier source verification architecture, a dual-model natural language processing prediction engine, a PIB fact-check cross-reference module, and a continuous incremental retraining pipeline. The three-tier architecture classifies news sources as Government Approved (Tier 1, anchored by Press Information Bureau domain whitelisting), Trusted Media (Tier 2, comprising DAVP-registered and internationally recognized publishers), or Unknown (Tier 3, requiring ML inference). Tier 3 articles are routed to either a fine-tuned MuRIL model for Indian-language and code-mixed content, or a fine-tuned BERT model for English and international content, each producing REAL, FAKE, or UNCERTAIN verdicts with normalized confidence scores. All articles are additionally cross-referenced against the PIB official fact-check archive, with confirmed matches receiving a PIB DEBUNKED override verdict. Classified verdicts are displayed via a real-time web feed with color-coded badges, confidence meters, and an on-demand URL analysis interface. The platform incorporates a self-improving retraining pipeline that incrementally fine-tunes both language models upon accumulation of 100 new fact-check-derived labeled samples, with drift detection triggering out-of-cycle retraining when classification confidence distributions shift significantly from baseline."
Disclaimer: Curated by HT Syndication.
