Métodos de estruturação de passos de reprodução em bug reports.

Abstract

Bug ixing is a natural step in the life cycle of a software, from development to the user interaction phase, problems will arise. Users, in general, report these problems through reports, however it is common that the information in these reports is incomplete or poorly structured. Among this information, the steps for reproduction elucidate the sequence of actions that reproduce the error and are considered as one of the most important information in the report. However, it is common that such information is not present or is present in an unstructured way, making it diicult, for example, for the developer to ind the problem and generate automatic tests. Thus, an application is necessary to identify and structure the steps to reproduce a bug by extracting information from the body of the text of the reports, in order to provide data that can be used in the most diverse applications. This project proposes the development of machine learning models based on Natural Language Processing with the ability to identify and structure the steps for reproduction. In which, high eiciency was evidenced in the detection of reproduction steps in a report (F1 = 0.69), but low capacity to promote the structuring of report components, such as actors and actions.