RAG system for answering 1,521 French questions on Madagascar higher education statistics (IndabaX Madagascar 2025)
# RAG Challenge: Madagascar Education Statistics Q&A
**Purpose:**
Notebook implementing a **Retrieval-Augmented Generation (RAG)** system to answer **1,521 French questions** about Madagascar’s higher education statistics using official MESUPRES reports.
**Competition Context:**
* **Event:** IndabaX Madagascar 2025
* **Type:** Individual, Madagascar residents only
* **Goal:** Extract accurate data from tables, graphs, and text; generate contextually grounded answers.
**Notebook Highlights:**
* Semantic chunking of tables/graphs with descriptive text
* JSON output for structured data and metadata
* Table-aware retrieval and acronym expansion for precise queries
* Supports LLM-based answer generation with context grounding
**Impact:**
Facilitates **data-driven insights** into Madagascar’s higher education, supporting policy analysis, research, and transparency.
**Notes:**
* French proficiency recommended
* Prepared specifically for the RAG competition
* Focus on accurate numerical extraction and contextually grounded answers