MUMBAI, India, Aug. 24 -- Intellectual Property India has published a patent application (202621082032 A) filed by Symbiosis International Deemed University on July 02, 2026, for Hybrid Transformer-Cnn System And Method For Sentiment Analysis In Low-Resource Language Texts.
Inventors include Nishit Bohra M; Avantika R. Patil; Omkar H. Thorve; Dr. Aniket K. Shahade; and Dr. Priyanka Deshmukh.
The application for the patent was published on August 21, 2026, under issue no. 34/2026.
Abstract: ABSTRACT HYBRID TRANSFORMER-CNN SYSTEM AND METHOD FOR SENTIMENT ANALYSIS IN LOW-RESOURCE LANGUAGE TEXTS A hybrid sentiment analysis system (100) and method for performing sentiment analysis on low-resource language social media texts is disclosed. The system (100) comprises a preprocessing module (120) for cleaning noisy text while preserving sentiment-bearing features and native script characters, a tokenization module (130) with sub-word segmentation, a transformer encoder module (140) for generating contextualized token-level representations, a CNN feature extraction module (150) with a one- dimensional convolutional layer (152) for extracting local n-gram sentiment patterns, a multi-strategy pooling module (160) that fuses global max pooling (162), mean pooling (164), and CLS token representations (166) into a comprehensive feature vector, and a classification head module (170) with progressive dimensionality reduction for mapping to sentiment class probabilities. The hybrid architecture captures both global contextual semantics and fine-grained local sentiment cues for improved classification accuracy in low-resource language texts. [
Disclaimer: Curated by HT Syndication.
