Robinson Chatbot
Project start: 2025-01-31
Project description
The Robinson Chatbot project explores the implementation of Retrieval-Augmented Generation (RAG) techniques to create an intelligent conversational agent knowledgeable about "Robinson Crusoe." The system combines state-of-the-art language models from both OpenAI and Amazon Bedrock with advanced text retrieval methods to provide accurate, contextually relevant answers based on the novel's content. Through extensive experimentation with various chunking strategies, embedding models, and prompt engineering techniques, the project demonstrates how RAG architectures can effectively enhance LLM capabilities for domain-specific applications while minimizing hallucinations and improving factual accuracy.
Main functionalities
- Implementation of various text chunking methods (e.g. by paragraphs, fixed token size, by chapters) to optimize information retrieval
- Experimentation with different embedding models to create semantic vector representations
- Advanced similarity search using FAISS vector database for efficient information retrieval
- Comparison of performance between OpenAI models (gpt-4o and gpt-4o-mini) and Amazon Bedrock models (Amazon Titan Text Express V1 and Amazon Titan Text Embeddings E1)
- Prompt engineering techniques to optimize context utilization and response quality
- Interactive Streamlit application for user-friendly chatbot interaction about Robinson Crusoe
- Comprehensive evaluation framework to measure accuracy, relevance, and coherence of responses
Skills
- Python
- OpenAI
- Amazon Bedrock
- AWS
- Numpy
- Faiss
- RAG
- Boto3
- Nltk
- Streamlit
- Prompt Engineerning
- Embeddings
- LaTeX
Project Report
You can download and review the complete project report with detailed methodology and results here: Robinson Chatbot - RAG Implementation Report
Sample photos
