Tamil AI Terminology Repository: Community Practice for Building Non-English AI Knowledge Systems

This post introduces a community-driven Tamil AI terminology project, which contains over 300 AI/ML terms organized in a four-column format (English terms, primary Tamil terms, alternative Tamil terms, and annotations). It aims to protect and develop non-English technical language resources in the AI era and break down language barriers in the dissemination of technical knowledge. The project is maintained by kpassoubady, open-sourced on GitHub, and was released on May 31, 2026.

Global AI resources are dominated by English, leading to unequal dissemination of technical knowledge and limiting learning opportunities for non-native English speakers. Tamil, a language with a long history and 80 million speakers, faces the dilemma of "vocabulary vacuum" in technical terms. This project (தமிழ் AI கலைச்சொற்கள்) aims to fill this gap, establish a localized AI concept expression system, and balance linguistic purity with technical practicality.

The terminology repository uses a four-column format:

1. English Term: Internationally accepted standard expression
2. Primary Tamil Term: Preferred translation approved by community discussions and experts
3. Alternative Tamil Terms: Synonymous or near-synonymous expressions
4. Annotations and Explanations: Definitions, etymology, usage scenarios, and translation considerations Currently, it contains over 300 AI/ML entries covering basic to advanced concepts (e.g., machine learning, attention mechanism, etc.).

The core principle of the project is to prioritize the use of pure Tamil vocabulary, such as using "நரவலை" (naravaḷai, neural network) and "சொல்துண்டு" (soltuṇṭu, token) instead of transliteration. These terms follow Tamil sandhi rules and compound word construction traditions. At the same time, it remains pragmatic: if an English term is widely accepted and there is no suitable Tamil alternative, the foreign term is retained and its status is noted.

The project adopts an open-source collaboration model and welcomes participation from various stakeholders. The quality control system includes:

• Reference Sources: Facebook's "சொல்லாய்வு குழு" (Vocabulary Research Group) and Anna University's 1998 "Computing Terminology Glossary"
• Version Management: Iterated to the third edition, optimizing consistency, annotations, and formatting
• Deviation Tracking: Maintaining a deviation document that records differences from authoritative recommendations and their reasons

The terminology repository is maintained in Markdown format, with the main file being ai-tamil-glossary.md and the reference document directory docs-glossary/. It is licensed under CC-BY-4.0, allowing free use, sharing, and adaptation (with attribution required). Communication channels include the Google Group (tamil-kalaisol@googlegroups.com) and Facebook community.

Implications of this project for the global AI community:

1. Linguistic diversity is the foundation of technical effectiveness (multilingual terminology helps AI serve global users)
2. Open-source communities have significant advantages in language standardization (rapid response, wide participation)
3. Ancient languages can express cutting-edge technical concepts Future plans: Expand terminology coverage to new AI concepts, simplify definitions, enhance linguistic purity, maintain format consistency, and track deviations.