MUMBAI, India, Aug. 24 -- Intellectual Property India has published a patent application (202641098829 A) filed by Ms Aadya Ranjan; and Dr G Devadasu on August 14, 2026, for System And Method For Incremental Knowledge Graph Distillation Using Temporal Consistency Constraints In Neural Models.
Inventors include Dr. T. Bhaskar; Sana Afreen; Mr. M. Saikiran; Dr. Anil Kumar; Mr. K. Sudhakar Reddy; and Mr. D Ranadeep Reddy.
The application for the patent was published on August 21, 2026, under issue no. 34/2026.
Abstract: A method (300) for incremental distillation of a knowledge graph (104) from a neural model is disclosed. The method includes maintaining, in a memory (204), the knowledge graph comprising a plurality of nodes representing entities and a plurality of edges. The method includes receiving a first output, and extracting from the first output a first set of candidate entity-relationship triples. The method includes receiving a second output and extracting a second set of candidate entity-relationship triples. The method includes determining a temporal consistency score, The method includes selectively updating the knowledge graph. The method includes inserting the candidate entity-relationship triple into the knowledge graph. The method includes revising the temporal validity interval of a corresponding existing edge. The method includes suppressing insertion of the candidate entity-relationship triple. The method includes generating a distilled subgraph, and providing the distilled subgraph to the neural model for use in generating a subsequent output.
Disclaimer: Curated by HT Syndication.
