SILVA, M. C.; http://lattes.cnpq.br/7973173797073048; SILVA, Melquisedeque Carvalho.
Abstract:
Product search is a fundamental functionality that allows users to locate and purchase specific items, and is applied in various contexts, such as e-commerce and price comparison websites. This study compares lexical and semantic approaches to perform this functionality. Although lexical search has advantages in terms of response time, it does not capture semantic relationships between words beyond lexical similarities. On the other hand, semantic search stands out by semantically capturing the relationship between terms, but in addition to being more complex, it can also be slower. In this work, we analyze lexical-semantic approaches for product data, specifically in two datasets: the Federal Government Materials Catalog and product descriptions present in invoices. We compare the search strategies considering the relevance of the results and the response time. The datasets have distinct characteristics, with the materials catalog being more formal and structured, while the invoices contain shorter and more informal texts, often with acronyms and abbreviations. This comparative study aims to identify trade-offs between lexical and semantic approaches, as well as to find the most appropriate strategies for each type of data. The results contribute to the selection of more effective search engines in product catalogs, considering different forms of data organization.
