MUMBAI, India, Sept. 28 -- Intellectual Property India has published a patent application (202641112909 A) filed by Dr. Ch. Rathan Kumar on September 21, 2026, for Leveraging Fine-Grained Sentiment Analysis And Graph Attention Networks For Review-Aware Recommendation Systems.
Inventors include Dr. Ch. Rathan Kumar; Dr. T. Sunil Kumar; Dr. Pillareddy Vamsheedhar Reddy; Dr. P. Bala Krishna; Mr. Gangadhar Shinde; Dr. Kakarlamudi V S Sudhakar; Dr. L. Vandana; Dr. M. Srinadh Swamy; Dr. M. Shailaja; Mrs. Bontha Mamatha; Mrs. Devarakonda Mounika; Dr. T Shyam Prasad; and Dr. T. Shekar Reddy.
The application for the patent was published on September 25, 2026, under issue no. 39/2026.
Abstract: The present invention discloses a review-aware recommendation system that leverages Fine-Grained Sentiment Analysis (FGSA) and Graph Attention Networks (GATs) to generate accurate, personalized, and context-aware recommendations. The system integrates Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), and Graph Neural Networks (GNNs) to analyze user-generated reviews, ratings, purchase history, browsing behavior, and product information collected from e-commerce platforms, online marketplaces, streaming services, and social networking applications. The invention performs fine-grained sentiment analysis to identify product or service aspects and determine aspect-level sentiment polarity, thereby capturing detailed user preferences beyond conventional document-level sentiment analysis. A heterogeneous graph representing users, items, reviews, and aspect relationships is constructed and processed using a Graph Attention Network, which assigns adaptive attention weights to neighboring nodes to learn meaningful semantic and structural representations. The sentiment features and graph embeddings are integrated within a recommendation engine to predict user preferences, rank candidate items, and generate personalized recommendations in real time. The system continuously updates its recommendation model using newly generated reviews and user interactions, thereby improving adaptability and recommendation quality. By combining aspect-level opinion mining with graph attention-based representation learning, the proposed invention enhances recommendation accuracy, addresses data sparsity and cold-start challenges, improves explainability, and provides scalable, intelligent recommendation services across diverse digital platforms.
Disclaimer: Curated by HT Syndication.
