MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202641083764 A) filed by Srinivas University Institute Of Engineering And Technology on July 08, 2026, for Eva – Enhanced Vernacular Assistant.
Inventors include Prof. Mahesh Kumar V B; Abhishek; Adarsh Kn; Vedavyasa Kalkura; Christy Jaison; V. Priyanka; Dhanush U. J; and Devathejas E.
The application for the patent was published on July 10, 2026, under issue no. 28/2026.
Abstract: ABSTRACT EVA – ENHANCED VERNACULAR ASSISTANT Neural Machine Translation (NMT) represents the cutting edge of linguistic technology, yet its application for under-resourced Indic languages like Malayalam and Kannada remains a significant challenge. This project develops "EVA" (Enhanced Vernacular Assistant), a robust, offline NMT system designed to bridge the linguistic gap through deep learning. Operating entirely within local system constraints to ensure data privacy and reliability, EVA leverages sequence-to-sequence transformer architectures for high-precision translation. The system integrates multi-modal inputs, including real-time voice recognition, to enhance accessibility. A unique feature of the project is the "Teacher Mode," which provides pedagogical breakdowns of translated sentences, making it an educational tool as well as a utility. Our research addresses the critical "Out-of-Memory" (OOM) errors common in local deep learning deployments by optimizing tensor loading and adopting memory- efficient MarianMT models (Helsinki-NLP) alongside traditional LSTM fallback mechanisms. The result is a high-performance translation ecosystem that preserves cultural nuance and linguistic structure without reliance on external cloud APIs. The system supports two target languages - Malayalam and Kannada - with an extensible architecture that can accommodate additional Indic languages. The Streamlit-based premium user interface features glassmorphism design, animated gradients, and responsive layouts that create an engaging user experience. Voice input capabilities using the SpeechRecognition library enable hands-free operation, while the integrated Text-to-Speech module allows users to hear translations spoken aloud.
Disclaimer: Curated by HT Syndication.
