Unificação de avaliações de produtos em E-commerce: uma solução computacional para apoiar decisões de compra online

Abstract

This study aimed to develop a web platform designed for the collection, integration, and unified presentation of product reviews from e-commerce platforms, focusing on Amazon and Mercado Livre. The research was motivated by the problem of fragmented online reviews, which makes product comparison more difficult and turns the consumer decision-making process into a more time-consuming and dispersed task. To address this issue, the study adopted an applied and experimental approach, structured in stages of automated data extraction, processing and standardization of the collected information, cross-platform integration, and the development of a web application for comparative visualization of results. Data collection was carried out through web scraping techniques using the Selenium, BeautifulSoup, and Pandas libraries, while the matching of equivalent products across the two platforms was performed based on Natural Language Processing techniques, using TF-IDF and cosine similarity. As a result, a functional platform was developed, capable of consolidating reviews and comparative information for 124 best-selling products from Amazon and Mercado Livre, gathering 16.568 collected reviews. The application allows users to visualize, within a single environment, data such as price, overall rating, number of reviews, and comments, as well as automatically identify the platform offering the best price and present summaries generated by artificial intelligence based on the collected comments, thereby contributing to a simpler and more efficient search and comparison process for consumers.