Design and Development of a Reusable Chatbot for Fashion Recommendations: Sinhala Language as a Case Study

Abstract

People communicate about their lifestyles, ideas, and social standing through fashion. Due to the ever-evolving desires and personal preferences and the availability of a wide range of options, selecting the ideal fashion design becomes difficult for customers. As a result, recommendation systems and chatbots have come to the market. In Sri Lanka, these chatbots are only available in English; thus, they are restricted in their ability to serve most Sinhalaspeaking community in Sri Lanka. To solve the said problem, this study builds a Sinhala fashion recommendation chatbot. This research focuses on the development of a chatbot developed using Python Flask. It uses OpenAI's GPT-3.5 Turbo to understand and generate human-like responses and Bing Translator to communicate in Sinhala language. The fashion problem solution was developed using the Soft Design Science Research Methodology (SDSRM). The chatbot was designed and developed using component-based software engineering (CBSE), which allows one to easily develop a complex application like a chatbot and export it to other domains. This makes it easy to deploy the chatbot for different purposes in business, education, or any other field with minimal modification. A hosted chatbot prototype was evaluated using a Google Form to obtain feedback. Users reported satisfaction with the personalized style suggestions, indicating the system's potential effectiveness. The unique achievement of this study was building a reusable chatbot module by combining several components; therefore, with minimum modifications, the chatbot can be customized to other domains as well.