NolBot is an AI-based support system designed to assist parents of children with developmental disabilities in daily communication and interaction.
The system leverages Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) to provide:
- Context-aware dialogue guidance
- Parenting skill recommendations
- Activity (play) suggestions
- Related educational video recommendations
Developed in Spring 2025 for the Business Big Data course.
The number of registered developmental disabilities has increased significantly in recent years, while access to continuous therapy remains limited.
Existing solutions face:
- High cost
- Regional imbalance in specialists
- Limited daily-life support
NolBot aims to bridge the gap between clinical therapy sessions by providing an AI-powered, low-cost, high-frequency daily support tool.
The system consists of multiple functional nodes:
- Dialogue Guide Generator
- Parenting Skill Information (RAG-based)
- Play Recommendation (RAG-based)
- Video Recommendation (LLM + filtering pipeline)
Key technical components:
- Prompt engineering
- Embedding-based document retrieval
- Chunking to overcome token limits
- Multi-stage recommendation filtering
- Token usage and cost tracking
- Categorizes conversation into planning, recall, and free discussion
- Encourages positive response strategies
- Avoids harmful reaction types (e.g. blame, correction overload)
- Uses embedded Q&A dataset derived from official parenting guides
- Contextual answer generation via RAG
- Generates activity suggestions based on video context
- Incorporates domain-specific educational materials
- Web-crawled database (~400+ videos)
- Two-stage filtering using LLM
- Token-efficient chunked processing
nolbot-llm-chatbot/
├── README.md
├── notebooks/
│ └── team5_nolbot.ipynb
├── slides/
│ └── nolbot-presentation.pdf- 박종휘
- 이은재
- 홍성연