Key points are not available for this paper at this time.
In the current digital era, efficiently accessing relevant information is crucial for various applications such as restaurant recommendation systems, website searches, book recommendations, among others. This study presents an approach for personalized recommendations in improving the customer service process in restaurants using OpenAI's Contextual Chatbot. The approach consists of six phases: (1) Data preprocessing, (2) Embedding and storage, (3) Scheduled updating, (4) Document retrieval, (5) Context adaptation and request creation, and (6) Response generation. OpenAI's text embeddings are used to convert application data into vectors and store them in a vector database. These vectors are used to retrieve similar records and generate contextualized responses using the GPT-3.5 model. The chatbot's performance is evaluated in terms of accuracy and user satisfaction. Two scenarios were used in the experimentation: (a) with the proposed solution and (b) without the solution. The results demonstrated an operational efficiency of 86.67% with the proposed solution and the versatility of the proposed methodology, showcasing its potential for application in a wide range of domains, including websites, books, PDFs, and other forms of documentation.
Romero et al. (Thu,) studied this question.