MUMBAI, India, June 26 -- Intellectual Property India has published a patent application (202641071779 A) filed by Sr University on June 09, 2026, for Knowledge-Enhanced Explainable Transformer Framework For Aspect-Level Sentiment Intelligence.
Inventors include Dr. Azizkhan F Pathan; Dr. Joydev Ghosh; Pooja Mv; Dr. B Vani; Santhosha P; and Swetha H U.
The application for the patent was published on June 19, 2026, under issue no. 25/2026.
Abstract: KNOWLEDGE-ENHANCED EXPLAINABLE TRANSFORMER FRAMEWORK FOR ASPECT-LEVEL SENTIMENT INTELLIGENCE The present invention discloses a Knowledge-Enhanced Explainable Transformer Framework for Aspect-Level Sentiment Intelligence. The system integrates aspect-relative positional encoding, sentiment-aware knowledge graphs, and explainability mechanisms into a unified architecture for fine-grained sentiment analysis. An Aspect-Relative Position-Aware Transformer (ARPAT) replaces global positional encodings with dynamic, aspect-anchored embeddings, enabling precise modeling of directional relationships between aspects and contextual opinion terms. A Sentiment-Aware Knowledge Graph (SAKG) encodes structured semantic relationships among aspects, opinion words, sentiment polarity, and contextual modifiers, which are processed by a Graph Neural Network (GNN) to generate knowledge-enriched aspect representations. A Hybrid Fusion Module adaptively combines transformer-derived contextual features with GNN-enriched structured features, producing robust sentiment predictions. An integrated Explainability Layer applies SHAP, LIME, and graph saliency methods to deliver token-level attribution, local surrogate explanations, and knowledge graph visualizations. The invention enhances accuracy, interpretability, and domain generalization, supporting deployment in healthcare, governance, and feedback monitoring systems.
Disclaimer: Curated by HT Syndication.
