MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202641085542 A) filed by Dr. Sowmya B J on July 13, 2026, for Development Of Hierarchical Software Requirements Specification Validation System With Ai- Enhanced Conflict Detection.
Inventors include Dr. Seema S; Dr. Geetha J; Dr. Sowmya B J; Mr. Armaan Pasha; Mr. Surya Sriram; Mr. Jeeth Kataria; Mr. Yashas N; Dr. Supreeth S; and Dr. Shruthi G.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: 7. ABSTRACT OF THE INVENTION The Patent disclosure covers Development Of Hierarchical Software Requirements Specification Validation System With Ai- Enhanced Conflict Detection. This invention presents an advanced Al-powered validation system for hierarchical Software Requirements Specification (SRS) documents, addressing critical challenges in large-scale requirements management across multi-level organizational structures. The system implements a nine-phase pipeline that processes requirements documents through sophisticated document ingestion using Docling for layout-aware parsing, structured field extraction using spaCy and Gemma 4 language models, hierarchy management with Doorstop for parent-child relationship tracking, knowledge base construction for value and predicate registry management, ’semantic’clustering wlth sentence-transformers and FAISS for'scope-aware requirement grouping, and S3CDA-enhanced conflict detection that validates entity-level contradictions to reduce false positives. Key innovations include the two-phase S3CDA methodology that achieves 40-60% reduction in false positive conflict detections by verifying that semantically similar requirements actually refer to the same domain entities before flagging conflicts, scope-aware clustering that only compares requirements within their validation chain rather than across unrelated document branches, automated change propagation with intelligent downward revalidation when parent requirements are modified, comprehensive error lifecycle management with auto-resolution capabilities when conflicts are naturally resolved by document updates, and entity-based conflict validation that distinguishes between true contradictions and coincidentally similar phrasing. The maintains complete traceability through integration with OpenFastTrace for coverage reporting. Unlike existing requirements management tools that rely on simple text comparison or manual review processes, this invention combines embeddings, large language model-based semantic extraction, and domain entity recognition in a unified platform that provides end-to-end validation from document upload through conflict resolution with minimal manual intervention. The system is designed for deployment in enterprise environments with SQLite/ PostgreSQL based persistent storage and supports version tracking.
Disclaimer: Curated by HT Syndication.
