Sistema de recomendação para apoiar a construção de gráficos com dados estatísticos

Abstract

Research on statistical data visualization emphasizes the need for systems that assist in decision-making and visual analysis. Having found problems in chart construction by novice users, we researched the following question: How can we support novice users to create efficient visualizations with statistical data? To address this question, this paper describes ViSC, a recommender system that supports the interactive construction of charts to visualize statistical data by offering a series of recommendations based on the selected data and on the user interaction with the tool. The system explores a visualization ontology to offer a set of graphs that help to answer information-based questions related to the current graph data. By traversing the recommended graphs through their related questions, the user implicitly acquires knowledge both on the domain and on visualization resources that better represent the domain concepts of interest. This paper also reports a qualitative study conducted to evaluate ViSC, using two methods: the Semiotic Inspection Method (SIM) and a Retrospective Communicability Evaluation (RCE) ---a combination of the Communicability Evaluation Method (CEM) and Retrospective Think Aloud Protocol. We first analyze how the questions influence the users' traversal through the graph and then address the broader question. We concluded the questions were important to generate efficient visualizations and thus, an efficient solution to help novice users in chart constructions.