Tsadra Dharma AI Project
In Collaboration With Mitra (Berkeley AI Research Lab)
Exploring AI Tools for Buddhist Translation
Welcome to Tsadra Foundation’s AI initiatives page. We’re working to support Buddhist studies through artificial intelligence tools and resources for scholars and translators. Through our collaboration with the Berkeley Mitra team, we’re developing innovative approaches that enhance research and translation while maintaining the integrity and depth of Buddhist teachings. Tsadra Foundation and Mitra are also engaged in a three-year collaboration project to develop AI tools specifically tailored for Buddhist translation and research.
You will find below all the webinars, slides, and case studies presented during our AI Tibetan Translation workshop.
Meet the Speakers
Professor Kurt Keutzer
Director, Dharmamitra
Kurt Keutzer is a Professor in the Berkeley AI Research (BAIR) Lab at University of California, Berkeley, where he works on making AI more computationally efficient. More generally, Kurt is part “Dharma center translator,” part amateur Buddhist scholar, and part AI researcher. Kurt brings all these together at Dharmamitra with the aim of making a deeper study of Tibetan and Sanskrit texts more accessible to practitioners and enabling scholars to fully harness the power of AI.
Sebastian Nehrdich
Technical Director, Dharmamitra
Sebastian specializes in natural language processing and computational linguistics at the Berkeley Dharmamitra team. His work centers on creating practical tools for Buddhist text analysis, with a focus on making these resources accessible to scholars and translators. He combines technical knowledge with an understanding of Buddhist studies to develop tools that serve the scholarly community.
Dr. Gregory Forgues
Director of Research, Tsadra Foundation
Greg brings together traditional Buddhist scholarship with digital approaches in his work. His academic background includes research at the Open Philology Project at Leiden University and doctoral studies completed in Vienna. His research encompasses Sanskrit, Classical Tibetan, and Classical Chinese, with particular attention to Mahāyāna sūtras, Tibetan tantric rituals, and Dzogchen teachings. Greg takes an integrated approach to text analysis, drawing on both traditional computational methods and newer AI technologies. His work incorporates corpus linguistics for analyzing patterns in Buddhist texts, text mining to extract key information, and semantic mapping to understand conceptual relationships. He uses these established methods alongside large language models, allowing each approach to complement the other.
Webinar Series
Webinar 1: The Dharmamitra AI Translation Project and AI Translation Tools
This video features presentations given during Tsadra annual meeting in Novemebr 2024. This session introduces the innovative work of the Dharmamitra Project, showcasing how AI technology is transforming Buddhist textual studies. It is followed by a discussion on how AI tools can be incorporated in the workflow of a translation project. Speakers emphasize that AI tools should be used carefully and critically. From their perspective, AI is not a replacement for human translators, but a valuable tool to assist translators in their work.
Through detailed demonstrations and expert insights, you’ll discover the powerful tools being developed for working with Sanskrit and Tibetan texts.
Detailed Timeline:
Introduction (0:00-5:50)
The video starts with introductions of the speakers and the organizations involved: the Shedra Research and Archive (SRA) and the Dharmamitra Project. The introduction also includes a brief overview of the potential of AI for Buddhist Studies.
- Overview of SRA and Dharmamitra Project
- Introduction to AI’s potential in Buddhist Studies
Kurt’s Presentation: Dharmamitra Project and Machine Translation (5:51-27:25)
Kurt introduces the Dharmamitra Project and its mission. He explains how they created the machine translation tools for Tibetan and Sanskrit, using aligned sentence pairs from existing translations. He also shares his experience using Dharmamitra for his translation work, showcasing examples of how it can be helpful and highlighting its current limitations.
- Project mission and scope
- Demonstration of translation tools for Tibetan and Sanskrit texts
- Current capabilities with aligned sentence pairs
- Real-world applications in translation work
Sebastian’s Technical Overview (27:30-42:30) Sebastian delves into the technical aspects of Dharmamitra, focusing on its features:
- Sanskrit machine translation system with grammatical analysis
- Combined translation and explanation system for Sanskrit texts
- Metra Search for Sanskrit metrical texts
- Dharma Nexus expansion beyond Buddha Nexus
- Chrome extension for seamless integration
Greg’s Practical Applications (42:40-1:31:00)
Greg’s presentation showcases how AI tools are useful to process large corpora of texts or to perform complex tasks under time constraints. Through several case studies he explains how AI tools like Google AI Studio and Claude can be used in practice to accomplish tasks that are part of the translation workflow:
- How to generate detailed Tibetan text outlines with Google AI Studio
- How to identify relevant texts in a corpus using Claude.ai
- How to verify Wylie transliteration and align Tibetan texts using Claude.ai
Conclusion and Discussion (1:31:00-1:45:00)
The video concludes with a brief discussion of the potential of AI tools for Buddhist studies and a call for collaboration between researchers, translators, and developers.
Webinar 2: AI Tools Workshop for Tibetan Translators (Part 1)
This hands-on workshop explores practical applications of AI in Tibetan translation workflows. Led by Dr. Gregory Forgues, the session demonstrates how AI tools can enhance translation projects while maintaining scholarly rigor and a respectful approach to Buddhist textual traditions.
Detailed Timeline:
Introduction and Overview (0:00-5:00)
- Welcome and introduction of speakers
- Overview of workshop goals
- Introduction to AI tools in translation
General Presentation: AI and Translation Workflow (6:05-56:45)
This part of the video features a slideshow presentation by Greg. He discusses the different stages of a Tibetan translation project and shows how AI tools can be used to support each stage. He also provides an overview of the evolution of computational methods for text analysis, including corpus linguistics, text mining, and semantic mapping. Greg explains how LLMs work and discusses key concepts such as tokens and context windows.
- Evolution of computational methods
- Understanding LLMs and their role
- Key concepts in AI-assisted translation
- Corpus linguistics and text mining fundamentals
- Semantic mapping techniques
Hands-on Demonstrations (56:45-1:47:30)
In this part of the webinar, Greg showcases various tools in relation to different textual environments.
- How to explore and summarize large Tibetan texts (i.e., Longchenpa’s Theg mchog mdzod, Atiśa’s Commentary on the Bodhicaryāvatāra) using Gemini in Google AI Studio to generate detailed outlines and structural analysis
- How to identify relevant texts from large corpora (i.e., Kyotön Tshultrim Mönlam’s collected works) using Claude.ai and a KWIC table as a dataset
Conclusion and Future Directions (1:47:31-1:51:41)
- Summary of key techniques
- Discussion of AI benefits for Buddhist studies
- Preview of advanced applications
Webinar Presentations
Presentation #1
Presentation #2
Webinar 3: Prompt Engineering for Buddhist Studies (Part 2)
This advanced session focuses on the art and science of prompt engineering, demonstrating how to craft effective instructions for AI models to enhance Buddhist studies research. In this video, Greg explains the key principles of effective prompting and gives an overview of prompting techniques, such as role assignment, multi-shot prompting, and active prompting. He emphasizes that prompting is an interactive (and iterative) process that requires experimentation, refinement, and human judgment. The presentation then shifts to practical case studies, showcasing how these prompting techniques can be applied in real-world Tibetan translation scenarios using AI tools like Claude, ChatGPT, NotebookLM, Perplexity, Gemini:
Detailed Timeline:
General Presentation: The Art of Mindful Prompting—Guiding AI Tools in Tibetan Text Translation (0:00-38:00)
- Welcome and Context Setting
- Translation Workflow Overview
- LLMs and Prompting Fundamentals
- Long Text Processing Strategies
- Active Prompting Methodology
- Advanced XML Tag Implementation
- Community Collaboration Frameworks
- Addressing participant questions about prompting techniques
Comprehensive Case Studies (38:05-1:34:00)
- How to compare different editions of canonical texts using Claude.ai
- Application: Comparing Derge and Narthang Tengyur catalogs
- Focus: Identifying commentaries on the Bodhicaryāvatāra in both Tengyur editions and comparing them
- How to verify and correct Tibetan Wylie transliterations using Claude.ai
- Application: Analyzing Gesar sādhana texts
- Focus: Error identification and correction in Wylie Tibetan
- How to create critical editions from multiple witness texts using Claude.ai
- Application: Generating textual alignments to produce critical editions
- Source: 28 Acintyabuddhaviśayanirdeśa witnesses from various canonical collections
- How to quickly compare texts with identical titles using Claude.ai
- Application: Rapid comparison of different texts entitled Bodhicittavivaraṇa
- Focus: Identifying textual variations and similarities
- How to conduct comprehensive research on Buddhist topics, traditions, doctrines, practices using Perplexity
- Application: Analysis of Chöd practice through available scholarship
- Sources: Sarah Harding’s publications on Chöd
- How to extract key insights from PDFs, video lectures on complex topics using NotebookLM
- Application: Summarizing educational content in the form of interactive podcasts, Q&As, and summaries
- Source: Sarah Harding’s lecture on Chöd history
- How to read Japanese, German, or French scholarship
- Application: Summarizing a Japanese paper on the transmission of the Bodhicaryāvatāra in the English language
- Source: Fujita 1993
- How to analyze historical Buddhist catalogs using ChatGPT
- Application: Analysis of St. Petersburg Tibetan text collection including 3,800 items
- Focus: Search of witnesses of the Saṃdhinirmocana Sūtra
- How to identify and evaluate online Buddhist resources using NotebookLM
- Application: Assessment of Buddha nature study materials from Tsadra’s Buddha nature website
- Focus: Using website data as input for summarization and Q&As
- How to discover and analyze academic literature using Perplexity
- Application: Finding relevant papers and monographs
- Focus: Mining Buddhist studies research materials such as open access PDFs
- How to create detailed lecture summaries from Youtube using NotebookLM
- Application: Processing educational content such as Dharma teachings, academic lectures available online
- Focus: Synthesis and analysis of lecture materials
Extended Q&A and Closing Discussion (1:34:00-1:53:55)
- Advanced implementation strategies
- Community feedback and suggestions
- Future developments in AI-assisted Buddhist studies
Webinar Presentations
Presentation #1
Webinar 4: AI Tools Workshop for Tibetan Translators (Part 3)
In this session session, we continue our exploration of AI-assisted translation with advanced prompting strategies and a selection of practical case studies.
Introduction (1.16-5:50)
- Collaboration is key to developing new tools and better prompting strategies
- Join us on Discord!
General Presentation: The Four Noble Prompts (5:52-20:22)
- Agentic Prompting
- Algorithmic Prompting
- Chain of Thought
- Tree of Thought
Practical Case Studies (20:23-1:26:00)
- How to identify entities (e.g., persons, places, terms, etc.) in Tibetan texts with a simple prompt
- Application: Kyotön Mönlam Tsultrim’s Commentary on the BCA
- Tool: Claude.ai (Sonnet 3.5)
- How to identify entities (e.g., persons, places, terms, etc.) in Tibetan texts with algorithmic and agentic prompts
- Application: Saṃdhinirmocana Sūtra
- Tool: Claude.ai (Sonnet 3.5)
- How to explore corpus content and topics of multiple texts using text mining tools (e.g., Wordtrees) and LLMs
- Application: Longchenpa’s Collected Works
- Tools: Wordtree script + Claude.ai (Sonnet 3.5)
- How to explore corpus content and topics of multiple texts using text mining tools (e.g., semantic mapping/word embeddings) and LLMs
- Application: Longchenpa’s Collected Works
- Tools: semantic closest neighbor + Claude.ai (Sonnet 3.5)
- How to understand semantic and logical relationships in a doctrinal discourse using “reasoning” models
- Application: Abhidharma texts and central treatises from various Buddhist traditions (Mahāvibhāṣa Śāstra,Vasubandhu’s Abhidharmakośa, Śāriputrābhidharma, Dhammapada, Visuddhimagga, Saṅgītipakaraṇa, Vibhaṅga, Pañcappakaraṇa)
- Tools: chatGPT o1
These case studies show how AI tools can enhance the quality and efficiency of translation work while ensuring accuracy and consistency in the final output. As in our previous sessions, we worked with diverse genres of Tibetan literature across a variety of Buddhist traditions and their related texts such as sūtras, śāstras, commentaries, pith instructions, etc. We used various AI-LLM tools including Claude’s Sonnet 3.5 and Open.ai’s o1 reasoning model.
Webinar Presentations
Webinar 4: AI Tools Workshop for Tibetan Translators (Part 4)
In this session session, we continue our exploration of AI-assisted translation with advanced prompting strategies and a selection of practical case studies.
Introduction to AI Translations Tools (0:00-38:32)
- Dharmamitra
- Monlam.ai
- DeepSeek v.3
- Claude.ai (Sonnet 3.5)
Practical application: All you want to know on how to translate a Tibetan text using an AI tool
Text used to demo the tools: Incipit of Mipham Rinpoche’s Dorje Tshegyal Gesar practice.
General Presentation: The Four Noble Prompts – CoT & ToT (38:33-42:51)
- Chain of Thought (CoT)
- Tree of Thought (ToT)
Practical Case Studies (42:59-1:26:32)
- How to review translations and develop translation quality assurance techniques
- Focus: Analysis of technical terminology (e.g., Sanskrit compounds translated into Tibetan)
- Application: Saṃdhinirmocana Sūtra
- Tool: Claude.ai (Sonnet 3.5)
- How to use a Tree of Thought prompt to extract the best of several models
- Focus: Comparison of several AI produced translations and translation improvement through prompts
- Application: Udānavargavivaraṇa
- Tool: Claude.ai (Sonnet 3.5)
- How to use a Network of Thought prompt to produce a translation from a complex text
- Focus: Translation generation through complex prompting
- Project documentation: Corpus-linguistic datasets (i.e., KWIC tables) and glossaries (Buddhist Hybrid Sanskrit terminology)
- Application: Udānavargavivaraṇa
- Tool: Claude.ai (Sonnet 3.5)
- How to annotate and index a translated text
- Focus: Identifying where footnotes can be added in a translation and defining index entries
- Project documentation: Translation materials ready for publication
- Application: Mipham’s De kho na nyid gsal ba’i sgron me, a commentary on Mañjuśrīmitra’s rDo la gser zhun
- Tool: Claude.ai (Sonnet 3.5)
These case studies provide step-by-step walkthroughs of AI tools across diverse Buddhist textual genres. Using various models and structured prompting techniques, we show how to enhance translation work while maintaining scholarly rigor.
Be Part of Our Growing Community
The intersection of AI and Buddhist studies is an exciting frontier, and we’d love for you to be part of this journey. Our community brings together scholars, translators, and technology enthusiasts who share a passion for preserving and understanding Buddhist teachings.
Begin your journey by exploring our comprehensive webinar series, where you’ll find practical demonstrations and insights you can apply to your own work. These sessions are just the starting point – we regularly host workshops, discussions, and collaborative sessions where you can connect with fellow practitioners and share your experiences.
Want to stay connected? Subscribe to our newsletter for the latest updates on AI tools, upcoming events, and community insights. Join our mailing list to receive invitations to future workshops and be the first to know about new resources and tools we’re developing. We hope to see you on our Discord channel Dharma Translation & AI Community to share new AI prompts and strategies.
Together, we’re creating new possibilities for Buddhist studies while honoring the depth and integrity of these precious teachings.