Crafting tools 💻 that make the Tamil language 📜 as sweet and accessible as nectar 🍯 in the digital landscape 💫
"தேமதுரத் தமிழோசை உலகமெலாம் பரவும் வகை செய்தல் வேண்டும்" (We must ensure the honey-sweet sound of Tamil spreads across the entire world.)
— Mahakavi Bharathiyar 🖋️
"யாதும் ஊரே யாவரும் கேளிர்" (To us, all towns are one, all men our kin.)
— Kaniyan Pungundranar 🌍
The bridge between ancient Tamil heritage and modern Artificial Intelligence is currently bottlenecked by poor data extraction. Most existing Tamil OCR (Optical Character Recognition) tools are built on sluggish Python wrappers, which struggle with large-scale processing. Furthermore, traditional OCR output for complex classical Tamil often results in heavily "noisy" text.
For fine-tuning Large Language Models (LLMs) on vast libraries of Tamil literature, speed and accuracy are non-negotiable.
Digital Tamizh was built to solve this:
- 🚀 Accelerate Data Extraction: Convert thousands of pages of Tamil PDFs and printed books into clean text datasets using Rust's multi-core parallelism.
- 🧠 Context-Aware Error Correction: Automatically repair misidentified classical characters using local Edge-AI.
- 🌿 Siddhar Literature Focus: Deeply optimized for the unique formatting and archaic vocabulary found in ancient Siddhar medical, yogic, and spiritual texts.
Explore our active projects bridging Tamil linguistics with modern computing:
| Repository & Link | Description | Primary Tech | Status |
|---|---|---|---|
| 👁️ tamil-simple-ocr | High-performance, parallel-processed OCR specifically designed for digitizing Tamil literature and accelerating LLM pipelines. (Accuracy: 97.3%) | 🦀 Rust 🤖 AI Models |
🟢 Active Build |
| 🕸️ digital-tamizh-website | The official web platform for our ecosystem, built with modern Island Architecture for lightning-fast, zero-JS delivery. | 🌌 Astro 📘 TypeScript |
🟢 Active Build |
| 🕷️ tamil-scraper-rs | A highly concurrent, blazingly fast web scraper for extracting vast Tamil text corpora for NLP and dataset curation. | 🦀 Rust |
🟢 Active Build |
| 🪷 thiruppugazh-sandhi-rs | A sophisticated rule-based engine to process, parse, and analyze the complex rhythmic structures (Sandhi) of Arunagirinathar's works. | 🦀 Rust |
🟡 In Dev |
We believe that preserving an ancient language requires the most advanced, bleeding-edge technology. Built with modern web standards, our architecture delivers maximum performance and zero bottlenecks:
| Technology | Purpose in Ecosystem |
|---|---|
| 🦀 Rust | Powers our core OCR pipeline, scrapers, and Sandhi engines for uncompromised multi-core parallelism and speed. |
| 🤖 Ollama + Gemma | Local Edge-AI models deployed for context-aware, semantic error correction. |
| 🚀 Tauri | Framework for shipping our high-performance, cross-platform desktop applications. |
| 🌌 Astro & TypeScript | Powers our frontend island architecture, generating static HTML with zero-JS by default for lightning-fast loads. |
| ⚛️ React 19 & Compiler | Drives dynamic UI islands with auto-memoization, eliminating manual boilerplate. |
| 🎨 Tailwind CSS | Utility-first styling for beautiful, culturally-themed, and responsive designs. |
Whether you are a Rustacean, an AI/NLP researcher, a Frontend wizard, or a Tamil scholar, your contribution matters!
- Star the Repositories: Show your support by starring projects like
tamil-simple-ocr. - Report Issues: Help us improve by reporting bugs or noisy text outputs.
- Contribute Code: Check out our
good first issuetags across the organization's repositories.
🌐 digital-tamizh.web.app | 🐙 github.com/digital-tamil
Made with ❤️ for Tamil. Built for the Future.