Vietnamese customer service chatbot for dental practice using large language models | |
| Author | Ngo Thanh Tinh |
| Call Number | AIT Project no.PMDS-23-09 |
| Subject(s) | Chatbots Dentistry--Data processing Natural language processing (Computer science) |
| Note | A project study submitted in partial fulfillment of the requirements for the degree of Professional Master in Data Science and Artificial Intelligence Applications |
| Publisher | Asian Institute of Technology |
| Abstract | Nowadays, Chatbots are used widely in various industries, in Dental care service, chatbots play an important role in enhancing patient satisfaction and optimizing the dental practice's productivity. Since the various benefits of chatbots for both dental practices and patients, It is crucial to develop and integrate a chatbot into the dental practice’s communication platforms and make them easily accessible to customers. Large Language Models (LLMs) offer a novel approach to developing chatbots that have the ability to not only effectively inquire people for specified information but also converse naturally and flexibly. While this development is exciting, such models do have downsides. In this study, I propose an approach to building an efficient chatbot based on the Retrieval-based language model to address the downsides of the chatbot based on only LLMs. I also experimented to compare different models: gpt-3.5-turbo, gpt-3.5-turbo-1106, vietcuna-3b-v2, and vietcuna-7b-v2 for developing chatbots in the dental care domain and utilize OpenAI's GPT-4 model as a judge to evaluate the selected models’ answers on a Benchmark dataset, that is created based on the consultation of an expert in the dental care domain, across six metrics Correctness, Comprehensiveness, Readability, Safety, Dentistry Literacy, and Overall Quality. The results show that with the proposed approach OpenAI’s model such as gpt-3.5-turbo and gpt-3.5-turbo-1106 demonstrates strong capabilities in handling questions related to the dental care domain, but these models still have limitations such as slower responses during peak traffic times, and expensive to scale. On the other hand, free open-source models such as vietcuna-3b-v2 and vietcuna-7b-v2 show promise for chatbot development, but they require significant computational resources for fine tuning with additional data to enhance performance. |
| Year | 2023 |
| Type | Project |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Professional Master in Data Science and Artificial Intelligence Applications (PMDS) |
| Chairperson(s) | Chaklam Silpasuwanchai |
| Examination Committee(s) | Vatcharaporn Esichaikul;Chantri Polprasert |
| Scholarship Donor(s) | AIT Scholarships |
| Degree | Professional Master in Data Science and Artificial Intelligence Applications - Asian Institute of Technology, 2023 |