ABSTRACT
The maturity of the computational argumentation field, demonstrated with the first live debate between a machine and a human,1 triggers a demanding question: how can we build argumentation technologies that bring people together? We believe that an important part of the answer is to include the audience's beliefs into the process.
ABSTRACT
This paper presents a visual analytics system for exploring, analyzing and comparing argument structures in essay corpora. We provide an overview of the corpus by a list of ArguLines which represent the argument units of each essay by a sequence of glyphs. Each glyph encodes the stance, the depth and the relative position of an argument unit. The overview can be ordered in various ways to reveal patterns and outliers. Subsets of essays can be selected and analyzed in detail using the Argument Unit Occurrence Tree which aggregates the argument structures using hierarchical histograms. This hierarchical view facilitates the estimation of statistics and trends concerning the progression of the argumentation in the essays. It also provides insights into the commonalities and differences between selected subsets. The text view is the necessary textual basis to verify conclusions from the other views and the annotation process. Linking the views and interaction techniques for visual filtering, studying the evolution of stance within a subset of essays and scrutinizing the order of argumentative units enable a deep analysis of essay corpora. Our expert reviews confirmed the utility of the system and revealed detailed and previously unknown information about the argumentation in our sample corpus.
