A Word-Level Digital Reader of the Prasthanatrayi with Sankara's Bhasya: Corpus, Method, and an Open, Offline Reading Aid for the Advaita Vedanta Canon
Summary
A new word-level digital reader for the Advaita Vedānta canon, the Prasthānatrayī with Śaṅkara's bhāṣya, offers an open and fully offline resource for non-specialists and students. This reader makes "every" word clickable, providing its euphonic split, morphological analysis, and gloss in a pop-up. Functioning as a concordance, it enables searches on dictionary headwords to find all inflected, sandhi-hidden, and compound-internal occurrences across both root text and commentary. The resource encompasses thirteen commentarial units, totaling 2,971 verses/sūtras/sections and 36,881 analyzed root-text word-occurrences, alongside a global dictionary of 95,587 distinct commentarial surface forms. Its hybrid analysis pipeline integrates rule-based sandhi-viccheda, a 927,000-form inflectional lexicon, attested-corpus look-ups, and LLM-assisted analysis with an adversarial two-pass verification. A durable human-expert review loop ensures correction persistence. High-confidence analyses demonstrate over 99% agreement with an authoritative inflectional lexicon. The reader is a self-contained HTML file, requiring no server or network access.
Key takeaway
For Sanskrit scholars and students grappling with the Advaita Vedānta canon, this digital reader provides unprecedented word-level access, significantly easing the study of complex texts. You can now directly interrogate "every" word for its euphonic split, grammar, and gloss, and trace concepts via lemma-aware search. For digital humanities researchers, consider adopting its hybrid LLM-plus-human-review pipeline to build robust, scalable tools for other low-resource or classical languages, ensuring accuracy and expert oversight.
Key insights
A digital reader combines traditional Sanskrit linguistics with LLMs and human review to make complex texts accessible.
Principles
- Hybrid analysis combines deterministic and LLM methods.
- Adversarial verification enhances LLM output reliability.
- Durable human-in-the-loop corrections are cumulative.
Method
The method involves normalisation, deterministic linguistic analysis, LLM-assisted analysis with adversarial two-pass verification, aggregation with confidence labeling, and a durable human expert review loop for corrections.
In practice
- Click any word for detailed grammatical analysis.
- Search by dictionary headword for all inflected forms.
- Utilize the self-contained HTML file offline.
Topics
- Advaita Vedānta
- Sanskrit Philology
- Computational Linguistics
- Large Language Models
- Digital Humanities
- Open Educational Resources
Best for: AI Scientist, Research Scientist, Domain Expert
Related on AIssential
See Counsel's argued verdicts on the open AI decisions leaders are weighing →
Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.