Advancements in AI-Driven Customer Service

Chatbots have gained more attention for resolving straightforward queries by retrieving data from relevant knowledge bases. As the capabilities of these interactive platforms have advanced, they have been integrated into user behavior, with customers using them for insights into past purchases or more information on future acquisitions. In this context, this paper presents the Amazon Question Answering Agent. This tool is designed to answer users’ frequently asked questions regarding products in Amazon’s electronics section. This objective will be achieved by embedding users’ queries and a pre-existing dataset, derived from users’ past questions, into a unified embedding space and then ranking the results to ensure the retrieval of relevant answers for optimal user satisfaction. The research evaluates four major algorithms—BM25, Word2Vec, GloVe, and FastText—for their effectiveness in generating sentence embeddings. The results showed that BM25 method outperforms the others in terms of accuracy and time efficiency, which increases the usefulness of the suggested Amazon Question Answering Agent. The study contributes to advancing chatbot capabilities in efficiently addressing user inquiries within e-commerce, promising enhanced engagement and satisfaction.