Improving Online Argumentation Through Deep Learning

摘要

Critical thinking and reasoning are essential for making informed judgments. This can be especially important for the intelligence community and associated government agencies. In this paper, we explore methods to evaluate critical thinking and reasoning. We focus on a public opinion/argument-based website - Yourview. We introduce the Yourview platform and present the annotated Yourview dataset that was created following a pilot period that focused on collecting public opinion on a range of topics. We then propose a method to classify arguments and their components related to the comments in the Yourview dataset. We assess the influence of components of argumentation as the basis for critical thinking and subsequently score and visualize these relations. Building on this, we predict critical thinking scores for what makes a good argument using a multilayer perceptron (MLP). The results of these models help enhance reasoning and establishment of knowledge from persuasive texts.