This paper introduces a comprehensive framework for developing an artificial intelligence–assisted clinical decision support system (CDSS) rooted in classical Ayurvedic knowledge. The study addresses a fundamental limitation in existing computational approaches to Ayurveda, where most systems depend on questionnaires or limited clinical datasets that are not explicitly derived from authoritative texts. To overcome this gap, the proposed approach focuses on systematically transforming unstructured knowledge from classical Ayurvedic literature into a structured and machine-learning-compatible format.

The framework is built upon a two-level data modelling strategy. The first level captures canonical disease–dosha–symptom–treatment relationships directly extracted from classical texts, ensuring fidelity to traditional knowledge. The second level generates context-aware patient instances by introducing controlled variability in demographic and lifestyle attributes while preserving validated clinical labels. A five-stage computational pipeline is designed to operationalize this process, including text extraction, semantic segmentation, domain-specific entity recognition, rule-based validation, and dataset expansion.

Machine learning models trained on the resulting dataset demonstrate strong predictive capability for dosha and disease classification, while treatment recommendations are generated through a knowledge-based system that maintains traceability to source texts. The integration of these components into a web-based platform highlights the practical applicability of the framework. Overall, this work establishes a transparent and reproducible pathway for combining traditional Ayurvedic principles with modern artificial intelligence techniques.

Keywords— Ayurveda; Clinical Decision Support System; Dosha Prediction; Machine Learning; Knowledge Extraction; Text Mining