Enhancing Line Chatbot Experiences: Utilizing Thai Question-Answering Systems to Enrich Dialogue with Unpredictable Questions

Abstract

Chatbots have gained popularity in recent years as a means of engaging with people using natural language, especially Line Chatbot, which has 290,000 active chatbots in Thailand. With such a high number of active chatbots in Thailand alone, Line Chatbot has become a prominent player in the chatbot market. But the main problem with chatbot construction is processing unpredicted questions. Even if Dialogflow, as a natural language understanding tool, can assist us in finding users' intentions, it is not good enough. In this work, we intend to enhance Line chatbot experiences by utilizing the pre-trained Thai Question Answering model as the NLP service that talks with the chatbot via the RESTful API. The major task is to deal with the message that Dialogflow cannot find the user's intention. With this strategy, we can improve the chatbot's ability to answer unexpected or complex questions, offering more accurate and relevant responses. By incorporating the Thai Question Answering model, the chatbot may exploit its pre-trained expertise to better understand and reply to user queries, boosting the overall user experience. The results show that the system met the specified requirements and consistently provided users with meaningful information.