DigAI: A Chatbot Interface for Fashion Recommender Systems

Abstract

The fashion industry has undergone a significant transformation in recent years with the advent of information technology and the proliferation of digital platforms. This change has created an urgent need to effectively communicate with users and address their needs in a personalized and meaningful way. However, the massive size of fashion item catalogs and the explosive number of product combinations and customer preferences have led to a phenomenon known as the information overload problem, which tends to degrade customers’ online experience. To mitigate the effects of this problem and improve customers’ online experience, many fashion businesses have implemented Dialog Systems (DS) as a solution. These systems allow users to interact with a platform and resolve product queries by serving as an interface. However, the complexity of human language poses a significant challenge to the effectiveness and acceptance of these systems, particularly in task-oriented and contextlimited scenarios. The success of a DS in understanding a user’s intent directly impacts their experience with the system. For instance, a sophisticated but poorly performing DS may be worse than a much simpler solution (e.g., a Graphical User Interface (GUI)). As such, designing an efficient and reliable system is critical to delivering satisfactory user experiences. To address this challenge, this work aims to design, develop and evaluate a chatbot called DigAI that serves as an interface for a recommendation system that assists users in finding clothing. To evaluate the chatbot’s performance, usability, hedonic and pragmatic values, potential users in Brazil assessed their overall satisfaction with the chatbot’s effectiveness in providing personalized recommendations as compared to a simpler GUI. This evaluation contributes to the broader effort to improve and personalize the online fashion shopping experience, thereby enhancing customer satisfaction and driving business growth.