MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641096236 A) filed by Diva Senthilnathan on August 10, 2026, for Peer Rubric: An Adaptive Key-Point Mining Framework For Automated Formative Assessment.
Inventors include Mrs V Radha; Mrs. S. Vigneshwari; Ashika R J; Adwin Jeso S; Akash E; Arunmozhi B; and Bhuvankumar M.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: The limitations of traditional assessment systems lie in their static and one-size-fits-all nature, which often fails to provide students with formative, transparent, and scalable feedback. Learners typically receive numerical grades or generic comments that do not capture the nuances of their responses or help them understand how to improve. To address this gap, we propose PeerRubric, a computer-implemented framework designed to enhance the quality and scalability of academic evaluation. At its core, PeerRubric leverages large-scale historical grading datasets, consisting of student responses paired with corresponding marks and grader feedback. Through advanced rubric inference, it performs statistical alignment between the textual content of answers, the assigned scores, and the specific comments from assessors. This process enables the generation of a Question Signature and a Key-Point Vector, which together capture the essential conceptual elements that were consistently rewarded across multiple grading sessions. Once established, these knowledge structures are reused in future assessments to provide adaptive, context-aware feedback. When a learner submits a new response to either the same or a semantically similar question, PeerRubric performs semantic alignment between the student’s answer tokens and the pre-generated Key-Point Vector. The system then dynamically highlights which critical concepts were satisfied and which were missed, offering students not just a grade, but an understanding of why the grade was earned. In parallel, PeerRubric projects a formative score with confidence calibration, making the evaluation both predictive and transparent. Importantly, the system incorporates a feedback loop where teacher overrides and assessor corrections are reintegrated, refining the rubric inference engine over time and preventing drift. This continuous adaptation promotes cross-cohort generalization, ensuring relevance across diverse learners. By transforming grading into a “living rubric,” PeerRubric bridges the gap between automation and pedagogy, enabling assessments that are not only scalable and efficient but also learner-centric and transparent.
Disclaimer: Curated by HT Syndication.
