MUMBAI, India, Aug. 10 -- Intellectual Property India has published a patent application (202621071499 A) filed by Indian Institute Of Technology Bombay on June 09, 2026, for Knowledge-State-Gated Neuro-Symbolic System And Method For Adaptive Alignment Of Operational Records With Structured Reference Plans.
Inventors include Kumar, Nanduri Sudheer; and Delhi, Venkata S..
The application for the patent was published on July 31, 2026, under issue no. 31/2026.
Abstract: A computer-implemented alignment system aligns operational records with elements of a structured reference plan using an adaptive knowledge-governed neuro- symbolic process. A persistent knowledge store maintains plan-driven entities, relationships, vocabulary mappings, domain constraints, and admitted knowledge derived from prior accepted associations. A knowledge-state metric is derived from the accumulated state of the knowledge store and indicates alignment-validated knowledge available to the system. For an incoming operational record, a probabilistic alignment component generates a candidate association and a first confidence value, and a constraint validation component evaluates the candidate association against domain constraints to produce a constraint-consistency value. A fusion controller computes a fused alignment confidence using relative influence determined by the knowledge-state metric and applies a disagreement control when probabilistic and constraint-based values diverge. The fused confidence is compared with a lower acceptance threshold and a knowledge-admission threshold governed by the knowledge-state metric to selectively accept associations and admit new knowledge for later alignments.
Disclaimer: Curated by HT Syndication.
