MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202631091453 A) filed by Azymant Systems Private Limited on July 28, 2026, for An Ai-Based Diagnostic Educational Coaching System With Autonomous Misconception Detection, Lifecycle Management, Agentic Remediation Scheduling, Spaced Reassessment, And Emergent Taxonomy Growth.
Inventors include Vijoy Ananda Bhadra; and Manash Sarkar.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The invention discloses an AI-based diagnostic educational coaching system comprising eight interconnected components functioning as an autonomous feedback pipeline: (A) a taxonomy-constrained diagnostic probe engine using a large language model (LLM) where a subject-specific misconception taxonomy serves simultaneously as a generation constraint for probe creation and an output classification schema; (B) a multi-signal false positive prevention module that separates persistent conceptual misconceptions from careless errors via an error_type classification filter and a confirmation threshold filter; (C) an autonomous misconception lifecycle state machine managing six distinct progress tracks from detection through confirmed resolution via LLM-evaluated response metrics; (D) an agentic background scheduler that runs automatically to prioritize unresolved gaps, generate targeted LLM micro-sessions, and stage them for proactive delivery at a learner's next application launch event; (E) a spaced reassessment engine that confirms durable retention by silently embedding context-shifted transfer questions into regular sessions at 7, 21, and 60-day intervals; (F) a prerequisite dependency graph that runs transfer checks during instructional unit transitions and gates onward educational progression upon failure; (G) an emergent taxonomy growth engine that runs batch cluster analysis on unclassified error logs via an LLM to propose new taxonomy entries through a human-in-the-loop validation flow; and (H) an assessment proximity sweep module that monitors upcoming scheduled assessment events and triggers a consolidated pre-assessment review session prioritizing outstanding misconception tags. The collective execution of these modules generates a persistent per-learner Learning Fingerprint tracking cognitive states as a detailed relational map of named misconceptions, which is used to personalize all downstream LLM session generation.
Disclaimer: Curated by HT Syndication.
