Avaliando modelos de LLM para personalização de consultas e aumento de relevância no e-commerce.

Abstract

The ability to provide personalized and relevant search results in a highly competitive e-commerce environment is crucial for customer satisfaction and the success of online stores. In this work, we explore a method to enhance the search experience in e-commerce using deep learning models to personalize user queries and improve the relevance of the returned items. The presented machine learning model was designed as a proof of concept to assess its ability to understand the context and intention behind user search queries, and to intelligently adapt them before being submitted to the search engine. The model rewrites the original query to prioritize customer's interest products by uncovering the underlying user intention and search context. Additionally, we also propose a classifier model that is responsible for selecting rewritable queries before using the rewriter model. This approach allows search results to be improved to highlight products of interest, significantly improving the relevance and effectiveness of the search.