Domain-Specific Chatbot Development Using the Deep Learning-Based RASA Framework

Abstract

Conversational agents are actively gaining popularity in research because of their ability to imitate human responses in almost every domain. As there are many research enhancements in deep learning models, it becomes challenging to incorporate all these enhancements while developing a conversational agent. One of the main advantages of conversational agents is their ability to answer frequently asked queries without any human involvement and automatically generate the conversation’s story flow. In any educational institution, it becomes difficult for the teaching and non-teaching staff to answer all the students’ queries regarding the course, exam, and other information regarding their daily activities in the institute. Using the deep learning framework, we developed a chatbot to answer various questions related to the education domain, such as exam(timetable, venue) and course-related queries(course handout). The questions are answered by querying databases which can be updated via an administrator’s web browser. The system will first create intents for the use cases and entity recognition mechanisms after connecting the deep learning framework to the database using custom actions. We had created a user interface to allow updates to the database for exam timetable and course information via either file upload or a web page.