Support for interactive analysis of online discussions combining data mining techniques

Keywords: Online discussions, Topic modeling, Sentiment analysis.

With the popularization of the Internet, online discussions have enabled an intense and constant exchange of knowledge between users on the most varied topics. While the data produced in these discussions enable new ways to to do science, your manual analysis is not feasible. Data mining techniques adopted, based on the discovery of knowledge in (Knowledge Discovery in Databases), an iterative process and with the specialist involvement in the analyzed problem domain, which aims to extract useful knowledge of big data. Each of this process presents numerous choices, significantly impacting the final research results. However, the participation of the specialist in the entire process presents itself as a challenge in multidisciplinary analyzes, where the problem domain and researchers often do not belong to the Data Mining area. Proposed works involve the researcher predominantly in the final data analysis stage, being mining specialists are responsible for defining the parameters for the other stages. There is also the absence of proposals for methodological processes for the analyzes, limiting its rigor and reproducibility.

The present work presents the research and development of a data mining tool for analysis interactive online discussion, as well as its use process, with a focus on participation of the researcher user in all stages of the analysis. To guarantee the scientific rigor of the development, the Design Science Research methodology is used, where the knowledge and understanding of a problem domain and its solution are achieved through a designed artifact. The proposed use process for applying the developed tool uses systematic data analysis to build theories, coming from Grounded Data Research. Your effective contribution is assessed using the System Usability Scale method, applied to users in the exploration of a real online discussion. The evaluation has a high acceptance rate, indicating the relevance and utility of the tool and its process of use. It is also A case study was carried out applying the tool and its process to the analysis of a online discussion on neuroscience. The consistency of the subjects produced by the analysis community and its ability to respond satisfactorily to research demonstrated the applicability of the developed research, besides contributing with the creation of new scientific knowledge in the area. The results of this work were communicated scientifically at IEEE 20th International international events Conference on Information Reuse and Integration for Data Science and II Latin American Workshop on Computational Neuroscience.