MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611070106 A) filed by Sharda University on June 04, 2026, for Multimodal Sentiment Analysis Using Confidence-Weighted Modality Gating.
Inventors include Mahawar, Vivek Kumar; Khan, Akil Ahamad; Shakil, Alameen; and Narwaria, Mamta.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present disclosure provides a multimodal sentiment analysis system (100) and a method (300) for affective inference from a textual input, an audio signal, and a visual signal. The system (100) includes a text processing unit (210), an audio processing unit (212), and a visual processing unit (214) that generate a text vector, an audio vector, a visual vector, and a corresponding confidence score for each modality. A gating unit (218) normalizes the confidence scores into a set of gating weights and derives a fused vector therefrom, and a classification unit (220) processes the fused vector to derive a sentiment polarity output (108), an emotion category output (110), and an emotion intensity score (112). Unlike conventional multimodal arrangements that treat each modality with equal importance regardless of signal quality, the present arrangement suppresses a corrupted modality and preserves a reliable affective inference for customer analytics and adaptive e-learning applications.
Disclaimer: Curated by HT Syndication.
