Processing of computer algorithms for traceability identification in scientific research

Abstract

Considering that scientific research is an essential part of the development of new knowledge and a multidisciplinary, time-consuming and error-prone task, it must be conducted under verifiable conditions in order to contribute to safe decision-making. The aim is to extract quality information from scientific articles automatically, presenting reliable, traceable and safe knowledge. To this end, this study investigates the perspective of identifying traceability and reproducibility patterns, using algorithmic Natural Language Processing methods, to demonstrate the identification of information contained in scientific articles, regardless of the research area. Therefore, in this work, the languages ​​Naive Bayes (NB), Cosine Similarity, Bag of Words (BOW) and Neural Networks (RN) were used for this purpose. As a result, it was possible to identify nine traceability patterns in the articles analyzed and propose an Artificial Intelligence model using algorithms with a minimum accuracy of 70%, demonstrating the traceability and reproducibility of the scientific articles analyzed.