MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641087913 A) filed by Vellore Institute Of Technology on July 17, 2026, for Adaptive Semantic Theme-Based Video Segmentation And Short-Clip Generation System.
Inventors include Dr. R. Manjula; and Badiyani Vipsa Miteshbhai.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: ABSTRACT ADAPTIVE SEMANTIC THEME-BASED VIDEO SEGMENTATION AND SHORT-CLIP GENERATION SYSTEM A method (100) for adaptive semantic theme-based video segmentation and short-clip generation comprises receiving a long-form video input (102), decomposing the video into audio and visual streams (104), generating a time-aligned transcript via automatic speech recognition and sentence tokenization (106), and generating sentence embeddings using a transformer-based sentence embedding model (108). Semantic theme-based segmentation is performed by applying density-based clustering to group semantically related sentences into thematically coherent segments (110). A complexity score is computed based on embedding variance, lexical entropy, and segment duration (310), and segments are routed through an adaptive hybrid model selection engine (300) based on the complexity score (312). Multibase features are extracted across a plurality of feature modalities (116), feature weights are computed using softmax-based normalization (414), and features are aggregated to compute engagement, emotion, and viral potential scores (120). Segments are ranked and clips are extracted with metadata (126). (Fig. 1)
Disclaimer: Curated by HT Syndication.
