About the project
How it worksஎப்படி வேலை செய்கிறது
THAMIZHARIVU AI is a source-grounded knowledge discovery system for Tamil literature, culture and digital heritage — built for Aurex'26, Track 05.
- 1
Query understanding
Detects Tamil, English, Tanglish, transliteration and mixed script, then normalises intent — no fixed question lists.
- 2
Dynamic source discovery
Finds candidate webpages, books, e-books, PDFs, uploaded documents and digital archives per query.
- 3
Retrieval & chunking
Fetches source content, segments it into passages and prepares embeddings for semantic matching.
- 4
Hybrid ranking
Combines keyword and vector similarity to rank passages by genuine relevance, not popularity.
- 5
RAG grounding
The language model may only answer from retrieved passages, in simple modern Tamil and English.
- 6
Citation & verification
Each answer carries source title, type, link, relevance and supporting passage. Unverifiable claims are flagged.
Current build status
The interface layer is complete and clearly marks every simulated state. The LLM, RAG, embeddings, vector search, hybrid ranking, conversation memory and Tamil OCR services are integrated in the following phases — the component and route architecture is already shaped around them.