Atribuições de novas funcionalidades de pré-processamento e mineração de dados para ferramenta UUX-Posts

Abstract

The traditional system evaluation methods present limitations, specialy when it comes to the user’s spontaneity on the evaluation moment. With that, arise studies and methodologies to system evaluation with the point of overcome these limitations, the textual evaluation methodology called MALTU, that carries the purpose of evaluating systems as stated by the Usage Related Posts (PRU), mainly in social systems. According to data such as posts in social systems, they can overcome the traditional methods limitations, but according to these data are not trivial, requiring automated techniques. With that, MALTU offers a tool called UUX-Posts to support professional systems evaluators. This tool in its initial version, provides support to the extraction and classification of PRUs steps, although presents some limitations like the classification, for still use the boolean search method to classify PRU posts, and does not provide preprocessing techniques, important things to the classifier performance. In its new version, the tool aims to provide support for all steps of the MALTU methodolody, including the PRUs classification from the contribution of this paper. Thus, this paper proposes the implementation of preprocessing functionality and data mining for the UUX-Posts tool, together with the use of more sophisticated classification algorithms compared to the boolean search. For this reason, researches were done throughout the literature, with the goal of verifying reliable and open source solutions that could be used to classify PRUs. As stated by this study, were defined techniques and proposed functionalities to, in other moment, be integrated in the tool. With the implementation and results, were found that the preprocessing contributes to a better classification result, although its contribution was not so representative, could be better improved with the use of more robust classifiers.