Grant: Alliance Grants

With the recent advances in natural language processing and AI, chatbots and large language models (LLMs) have shown great promise in answering questions in various domains, including drug-related questions. However, prior work has shown that when it comes to answering specific questions about drugs, out-of-the-box chatbots (even chatbots that utilize LLMs) tend to fail to give complete information and/or answer complex questions. Hence, the goal of this project is to partner with Sandoz, a world leading drug manufacturer to develope specialized chatbots to answer drug-related questions. Specifically, we will build a taxonomy of Pharma/Drug related questions from existing question dataset with the help of teams at Sandoz Canada. Then, we will develop a highly accurate chatbot, using Natural Language Understanding, that can answer questions based on pre-defined intents. Third, we will use drug monographs and develop chatbots using LLMs to answer drug-related questions. To evaluate the effectiveness and accuracy of the chatbots, we plan to perform empirical studies that measure the chatbot's abilities in generating accurate answers, saving effort, reducing turnaround time, and ultimately, increasing customer satisfaction.