Uma Abordagem Baseada em Métricas de Redes Complexas para o Estabelecimento do Grau de Influência de Termos em Documentos

Abstract

In the last years, the information retrieval field has received much attention from the world scientific community. Research on the improvement of methods and algorithms for textual information retrieval has increased, largely concentrated in the improvement of vector model, especially in efficient methods and functions for similarity calculation between documents and queries. In parallel, the networks analysis subject has attracted the interest of the scientific community due to its ability to represent complex issues in an objective manner, offering a theoretical and practical approach for the study of the properties and behavior of the elements and relations of which problems are made. Recently, research papers considering documents as word complex networks has been developed. However, using this approach to solve information retrieval and classification problems has been under-exploited. This paper presents an approach, based on metrics of complex networks, that obtain functions to assign weights to terms in documents. The approach performs as well as a vector model based approach, when applied to estimate the similarity between documents and queries from a reference collection. This demonstrates the applicability of the metrics of word complex networks in information retrieval problems.