Metaciência da Computação: um estudo sobre taxonomia de pesquisa

Abstract

The problem in presenting or validating results of scientific publications has been studied for a long time, and the literature chronologically has been pointing out these pieces of evidence. Such an issue still persists in the present days. Recent evidence confirms this fact, indicating that something needs to be done to enhance scientific research in the field of computing. One of the reasons highlighted by the literature is the lack of knowledge about the research itself. The possibilities, nuances, and dif ferent shades that will allow its enrichment. Departing from traditional experimental research without effective validation of results. This paper presents a meta-study of Computer Science, from which a Research Taxonomy is proposed to guide research and teaching. The obtained results are the 11 steps that comprise the stages of sci entific research, specialized in the context of computer science, and its logical cycle of successive refinement. The points of insertion of computer research within science as a whole, these being Meta, Middle, and End. The proposed taxonomy systematizes the concepts of 96 types of classifications identified in the literature, grouped into 10 categories. The validation process of this taxonomy identified its applicability and com pleteness. In this process, an ontology was first developed based on the descriptors of the taxonomy, conducting a textual search in 2,508 scientific publications from 68 jour nals published between 2018 and Aug/2022. The 124,601 occurrences resulting from this research had their context analyzed and classified by the semi-supervised k-NN algorithm, seeking to identify whether the context of the descriptor’s usage conforms to the proposed taxonomy. The results indicate that this happened in 73.37% of the con texts of descriptor usage, demonstrating its applicability. To assess its completeness, the areas of the journals from which the publications were extracted were analyzed. This information follows the SCOPUS classification, and for this analysis, the ACM and SBC’s GAS classifications were also considered, for which an equivalence parallel was drawn. Within this exercise, it is possible to verify that all sub-areas of comput ing according to the three classifications are covered in the results, demonstrating its completeness concerning Computing.