| Summary: | Since its early days in 2003, Amazon Web Services (AWS) has evolved rapidly. From a single service created to support its parent company’s e-commerce business, AWS became a leading cloud services provider. As AWS‘s product offerings and customer base expanded, its support knowledge base grew proportionally. Customers looking for self-service support solutions need novel solutions to navigate such a vast repository of information. This study explores a set of knowledge retrieval architectures designed to surface the most relevant content to customers pursuing self-service solutions within the knowledge base of a large technology company. To recommend the best content that a customer should consume next in their journey, we leverage insights about the content already seen by the customer. Our research encompasses three methodologies: semantic search utilizing large language model embeddings, a frequency-based n-gram model, and a hybrid approach integrating semantic search within a deep neural network framework. Simulations on historical data display a significant percentage of scenarios where customers would be accurately directed to the desired solution. Our findings suggest that organizations can adopt these methodologies internally to enhance digital customer journeys and pave the way for further innovations in this domain. This study addresses the immediate challenges of navigating large-scale company knowledge bases and presents the potential for scalable self-service models.
|
|---|