US20250200627
2025-06-19
Physics
G06Q30/0627
The patent application discusses a method for enhancing interactions with an artificial intelligence virtual assistant using a large language model (LLM). The LLM is specifically trained to include detailed product information stored within a product knowledge base. This setup allows users to interact with the AI through an interface embedded in websites or applications, primarily for inquiries related to products available for sale.
Users engage with the AI by providing input related to the products they are interested in. The LLM processes this input and generates responses that are directly informed by the data in the product knowledge base. These responses are then transformed into video segments, which are presented back to the user, facilitating a dynamic and engaging interaction experience.
Following the presentation of video responses, users have the opportunity to provide additional input or feedback. This subsequent input is crucial as it is evaluated to initiate self-improvement processes within the system. The goal is to refine both the content of the knowledge base and the quality of responses generated for future interactions.
The self-improvement process is multifaceted, including several key components:
The system is designed to perpetually evolve by integrating user feedback and external data sources. This continuous enhancement ensures that the virtual assistant remains up-to-date with current product information and improves its ability to deliver accurate, relevant, and helpful responses, thereby boosting user satisfaction and engagement.