MUMBAI, India, Sept. 22 -- Intellectual Property India has published a patent application (202641109059 A) filed by Acharya Institute Of Technology on September 11, 2026, for System And Method For Outcome-Weighted Multimodal Custom Intent Embedding For Marketing Management.
Inventors include Mr. Arunmani; Mrs. Anju A Nair; Mr. Yogesh Dixit; Dr. Ramesh C Raina; Prof. Rashmi C. M; Prof. Shilpa N; and Prof. Ashwini P.
The application for the patent was published on September 18, 2026, under issue no. 38/2026.
Abstract: A computer-implemented system and method for outcome-weighted multimodal custom intent embedding for marketing management is disclosed. The system fuses text, image, behavioural, and transactional signals into a customer intent embedding compared against marketer-defined custom intent class prototypes computed from seed exemplars, permitting new intent classes without full retraining. Each prototype is associated with a confidence radius and supports negative seed exemplars. Downstream marketing outcomes are captured and converted into outcome weights using a specific formula combining outcome type scaling, directional sign, logarithmic magnitude scaling, and exponential recency decay. The weights are applied through a weighted contrastive loss to recalibrate a Cross-Modal Fusion Engine with modality-specific gating and the prototype vectors, while leaving modality-specific encoders unchanged, thereby reducing recalibration cost by approximately 69%. The recalibrated embeddings drive audience segmentation, next-best-action recommendation, creative matching, and marketing budget allocation within a closed feedback loop. Prototype drift is monitored and alerts are generated when drift exceeds a threshold, prompting seed exemplar review.
Disclaimer: Curated by HT Syndication.
