Intelligent Identification System for Sanskrit Prosody: A Classical Literature Analysis Tool Integrated with Explainable AI

This article introduces the Chandas-identification project, an open-source tool that uses machine learning technology to automatically identify Sanskrit poetic meters (Chandas). It supports 10 common meters and integrates SHAP explainability analysis, providing a modern technical solution for Sanskrit research and the digitization of classical literature.

Sanskrit carries thousands of years of philosophical, religious, and literary traditions. Its poetry is renowned for strict metrical rules (Chandas), which define syllabic stress patterns. Traditional identification requires profound Sanskrit knowledge and practical experience, is time-consuming and error-prone, especially for analyzing long texts.

Chandas-identification is a desktop application that supports the identification of 10 common meters. It provides real-time analysis, confidence scoring, alternative suggestions, and SHAP explainability visualization. The interface is simple, including an input area, analysis button, result display area, and explanation view, making it easy for users without technical backgrounds to use.

Feature engineering includes syllabic stress marking, positional encoding, and vowel-consonant patterns; the model may use RNN, CNN, Transformer, or ensemble methods; SHAP technology visualizes the contribution of each part of the text to the identification result, enhancing user trust and educational value.

For Sanskrit learners: As a digital tutor to accelerate prosody learning; for researchers: Batch text analysis to support literary style and dating studies; for digitization projects: Improve metadata annotation efficiency; for comparative literature: Provide quantitative analysis methods.

Facing issues such as input quality (errors affect identification), meter coverage (only 10 types), encoding compatibility (adapting to multiple standards), and API integration (improving documentation), suggestions include adding error correction functions, expanding meter support, enhancing encoding compatibility, and optimizing API services.

The project represents a new paradigm of AI in cultural heritage protection, realizing standardized and large-scale analysis and democratization of knowledge dissemination; it does not replace experts but helps inherit ancient cultural heritage, opening a new door for Sanskrit research.