Design of an Academic Services Chatbot at Asia Institute Malang

Abstract

This study successfully developed and evaluated an academic services chatbot utilizing Natural Language Processing (NLP) and Artificial Neural Networks (ANN), showcasing a comprehensive methodology that encompassed data collection, preprocessing, model training, and performance evaluation. We meticulously derived the dataset from frequently asked questions at the Asia Institute of Technology and Business Malang, ensuring its relevance and applicability to the target audience. Rigorous preprocessing techniques, including tokenization, stemming, and stop words removal, were employed to enhance the quality of the input data for the ANN model, which significantly improved its performance. The training results revealed a strong correlation between the number of training epochs and the accuracy of the chatbot's responses, indicating that increased training led to enhanced performance. Furthermore, a Cronbach's Alpha coefficient of 0.965 confirmed the validity and reliability of the measurement tool for user feedback, highlighting the robustness of the collected data. User testing involving 37 students indicated a high level of satisfaction with the chatbot's performance, as it achieved a perfect accuracy score of 100%. These findings highlight the potential of NLP-based chatbots to enhance academic information services, effectively addressing student inquiries while significantly reducing the workload on academic staff. This study serves as a valuable model for other educational institutions aiming to implement AI-powered solutions to improve their academic support services and overall student experience.