A Semantic Approach to Support the Analysis of Abstracts in a Bibliographical Review

Abstract

A large amount of scientific information is found in digital databases. This fact turns the search for the state of the art an increasingly challenging task. This work has the goal of developing a technological solution to support researchers in bibliographic review using the Latent Dirichlet Allocation. The Sw3T software, developed in Python, analyses the abstracts in scientific publications that interest the researcher. Through the use of semantic analysis, Sw3t supplies themes, which are defined by a group of terms. Themes can subsidize the researcher in the process of identification of complementary terms through the restriction on the number of publications of interest and it contributes to the analysis of the publications selected by their classification by theme as well. Sw3t has shown to be efficient and increased reliability in the search process regarding complementary terms, as well it contributes to the promptness of the detailed analysis of the publications. The results of the bibliographical review, regarding the application of Sw3T, have shown to be consistent and replicable. Future work will address the use of Sw3T for other bibliographic reviews as well as improvements and broadening of its functionalities such as search in DDB automation and previous content processing in abstracts.