MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641050024 A) filed by Srm Institute Of Science And Technology; Mr. S. Anirudh; Ms. Jeyashanchitha. J; Ms. Sakshi; and Dr. Sabitha. P, Chennai, Tamil Nadu, on April 20, for 'real-time sentiment classification device utilizing hierarchical fusion transformer architecture and contrastive learning.'
Inventor(s) include Mr. S. Anirudh; Ms. Jeyashanchitha. J; Ms. Sakshi; and Dr. Sabitha. P.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "This invention stems a highly developed, proprietary hardware system, creating a sophisticated interface to bridge deep learning and user input and output. At the heart of this device is the proprietary Hierarchical Fusion Transformer, enabling the user to converge multiple input streams, including an audio signal and a textual stream. Implementing contrastive learning for this device improves the system's emotional range discrimination and classification ability in real-world environments, ultimately improving accuracy and precision. The planned physical design of the device illustrates the balance between the "black-box" nature of the technology and the "clear-box" nature for accessibility for the user. For instance, the device will display in high resolution the emotional classification, along with the statistically determined confidence in that classification, in a cutom user-interface. An external signaling device will provide immediate feedback in the form of a color-coded system to signal optimally and quickly. The hardware is centered around a powerful core designed to perform high intensity tasks quickly and without a barrier to their processing speed. The device is designed such that continuous intensive processing is supported by a system that electronically and passively manages the airflow around the core to dissipate the heat from the components. This device demonstrates the newly developed capabilities in the area of affective computing in a compact, highly designed housing providing immediate operational emotional intelligence. The invention brings the sophisticated collapsed transformer logic, originally housed in the cloud, down to a local device. Not only does this provide a secure, real-time, low-latency basis for monitoring the emotional sentiment of individuals, it is advantageous for improving customer service, most immediately for the social robotic field and psychological studies. This device design embodies the aforementioned goals, ensuring that affective computing remains accessible and valuable."
Disclaimer: Curated by HT Syndication.
