MUMBAI, India, June 30 -- Intellectual Property India has published a patent application (202641073802 A) filed by Vellore Institute Of Technology on June 14, 2026, for Transformer-Based Semantic Plagiarism Detection Architecture Using Siamese Neural Network And Deep Code Embeddings.
Inventors include Thanuja; Dilipkumar S; Janhvi Mangesh Dixit; Aditya Prashar; and Mitisha Sachdeva.
The application for the patent was published on June 26, 2026, under issue no. 26/2026.
Abstract: The present invention relates to semantic plagiarism detection using transformer-based deep code embeddings and Siamese neural network architecture. The invention receives source code snippets and preprocesses the source code through normalization and tokenization operations. Semantic embedding vectors representing logical behavior and functional characteristics of the source code are generated using a transformer-based encoder. The semantic embeddings are processed through a Siamese neural comparison architecture employing shared encoder weights for semantic similarity evaluation. Cosine similarity computation and threshold-based classification operations generate automated plagiarism verdicts corresponding to plagiarized or non-plagiarized source code. The invention enables semantic plagiarism identification despite code obfuscation operations including variable renaming, formatting modification, and structural transformation. The invention further provides enhanced plagiarism detection accuracy, multi-language source code analysis capability, automated semantic comparison, and integration with educational assessment and software evaluation environments.
Disclaimer: Curated by HT Syndication.
