stem-content-ai-project/content-pipeline

Swahili Content Generation

This Hugging Face Space provides an AI-powered tool for generating educational content in Swahili for primary school students in grades 3 and 4. The system uses a Retrieval-Augmented Generation (RAG) approach to create accurate and contextually relevant educational materials.

Features

• —Generate educational content for Math and Science subjects
• —Support for grades 3 and 4
• —Three different content styles: normal, simple, and creative
• —RAG-based approach for accurate and contextually relevant content
• —Easy-to-use interface

How to Use

1. 1.Select the grade level (3 or 4)
2. 2.Choose the subject (Math or Science)
3. 3.Enter a topic (e.g., "namba", "mazingira", "vipimo", etc.)
4. 4.Select a style (normal, simple, or creative)
5. 5.Click "Generate Content"

Available Topics

Grade 3 Science

• —mazingira (Environment)
• —nishati (Energy)
• —maada (Matter)
• —mawasiliano (Communication)
• —usafi (Cleanliness)
• —vipimo (Measurements)
• —mlo (Nutrition)
• —mfumo (Systems)
• —maambukizi (Infections)
• —huduma (Services)
• —vifaa (Equipment)

Grade 4 Science

• —kinga (Immunity)
• —magonjwa (Diseases)
• —majaribio (Experiments)
• —maji (Water)
• —ukimwi (HIV/AIDS)
• —huduma (Services)
• —mazingira (Environment)
• —nishati (Energy)
• —mfumo (Systems)
• —mawasiliano (Communication)

Grade 3 Math

• —namba (Numbers)
• —mpangilio (Arrangement)
• —matendo (Operations)
• —sehemu (Fractions)
• —maumbo (Shapes)
• —vipimo (Measurements)
• —fedha (Money)
• —takwimu (Statistics)

Grade 4 Math

• —kugawanya (Division)
• —kujumlisha (Addition)
• —kuzidisha (Multiplication)
• —namba (Numbers)
• —kirumi (Roman Numerals)
• —wakati (Time)
• —mpangilio (Arrangement)
• —vipimo (Measurements)
• —takwimu (Statistics)
• —kutoa (Subtraction)
• —fedha (Money)
• —sehemu (Fractions)
• —maumbo (Shapes)

Technical Details

This application uses:

• —Meta's Llama-3.2-3B-Instruct model for text generation
• —FAISS for efficient vector similarity search
• —SentenceTransformers for text embeddings
• —FastAPI for the backend API
• —Gradio for the user interface

The system employs a Retrieval-Augmented Generation (RAG) approach, which enhances the language model's output by retrieving relevant information from a curated database of educational materials in Swahili.