MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202621106053 A) filed by Lnct University on September 03, 2026, for System And Method For Evidence-Verified Classification And Reproducible Assessment Of Candidate Documents.
Inventors include Dr. Krishna Kumar Thakur; Dr. Nandita Tiwari; Pushkar Gautam; and Shaila Batham.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: ABSTRACT SYSTEM AND METHOD FOR EVIDENCE-VERIFIED CLASSIFICATION AND REPRODUCIBLE ASSESSMENT OF CANDIDATE DOCUMENTS The present invention provides a computer-implemented system and method for evidence-verified classification and replicable evaluation of candidate documents. A document processing subsystem transforms each candidate document into a canonical sequence of characters with the preservation of source coordinates. An evidence graph subsystem presents the extracted candidate assertions by means of evidence nodes associated with verbatim source spans, cryptographic digests, confidence values, and verification status. A probabilistic inference subsystem computes multi-dimensional evaluation scores, whereas a deterministic eligibility subsystem uses chronological, provenance, dependency, and inter-record consistency predicates. Failed evidence nodes are isolated and only dependent score subcomponents are recalculated. A deterministic adjudication subsystem applies confidence calibration and versioned thresholds to assign a decision division. A machine-verifiable decision certificate defines the admitted and isolated evidence, models, rules, thresholds, and scores. The decisions are finalized before issuance to a write-once hash chained audit log. An independent decision replacer recreates the classification without resorting to the inference subsystem. An output digest associates the dossier with the path.
Disclaimer: Curated by HT Syndication.
