Ai Chatbot: Improve Efficiency in Handling Student Queries at the Department of Computing and Informatics, Nairobi University

Abstract

Artificial Intelligent chatbots are increasingly replacing human chat service agents because of bot’s ability to communicate with humans through natural language and artificial intelligence (AI) technologies. Studies have found that universities need to provide effective and efficient digital platforms, powered by AI, in order to support holistic virtual learning (Alharthi, Spichkova and Hamilton, 2019). The COVID-19 pandemic has disrupted face to face learning and most institutions have now adapted hybrid or pure online learning. In cases where new students are onboarded virtually, there has been an increase in student queries and the traditional channels of human support is becoming ineffective. This research aimed to develop a conversational AI Chabot that would improve efficiency in handling Student Queries at the Department of Computing and Informatics (DCI) at the University of Nairobi. Waterfall Software development methodology was used in this development study. The source of data was the content on the University of Nairobi website and DCI students. Content analysis and structured interviews were used to obtain the data. Natural Language Processing (NLP) and LSTM model were used to build the AI Chatbot (dubbed UniBot). BLUE evaluation method was used to assess the effectiveness of the UniBot in providing accurate responses. The research established a near perfect match in chatbot response with a BLUE score of 0.75. Quantitative approach was also adopted to evaluate the efficiency of the model by having 20 target students use the chatbot for a duration of 3 weeks and give their feedback through questionnaires. A mean score of 4.10 and standard deviation of 0.59 was obtained from the students’ responses, meaning; the UniBot achieved its objective of improving efficiency in handling students’ query. Particularly, students who interacted with the UniBot indicated that the bot was easy to use and could retrieve the needed information very fast. However, the chatbot was not able to answer questions on topics that it had not been exposed to and would leave the question unanswered. In such instances, it’s recommended that the chatbot should provide relevant links or human contacts.