CancerBot: A Retrieval-Augmented Generation based Cancer Chatbot Using Large Language Models

Abstract

Cancer is the second leading cause of death globally, highlighting the urgent need for innovative solutions to support patients and healthcare providers. Remote and digital cancer care interventions, such as chatbots, are more relevant than ever, offering timely information and personalized assistance. With the rise of conversational AI, chatbots have emerged as a potential solution to bridge this gap, offering 24/7 assistance and personalized responses. This paper introduces CancerBot, a Retrieval-Augmented Generation (RAG) system designed to address the need for reliable, contextually relevant cancer-related information. Built on top of a large language model LLaMA-2, and integrated with a vector database, CancerBot retrieves relevant medical literature to generate accurate responses, minimizing the risk of hallucinations. Through synthetic testing and evaluation using metrics such as faithfulness, context relevancy, and response relevancy, our results demonstrate that the RAG-based CancerBot significantly outperforms the standard LLaMA-2 model in addressing cancer-specific queries. This advancement showcases the potential of RAG systems to improve patient care by providing reliable, context-driven information that enhances patient engagement and healthcare decision-making.

Ähnliche Arbeiten