Can Deliberation be Detected Computationally?

One major challenge in measuring deliberation is scalability: it requires extensive manual analysis to classify texts based on different characteristics, like respect to other speakers, the level of justification in argument etc. However, recently computational text analysis has been successfully applied into understanding which linguistic characteristics define pleasantness as well as helpfulness when answering questions. Recently, political scientists have started to examine computational analysis tools for text, and observed that the results with these tools align with expectations. For classifying deliberation, we apply supervised machine learning approach. In this approach, we have labeled certain texts based on their deliberativeness. From the text, we extract certain linguistic characteristics to create a feature vector, and using these analyze which linguistic characteristics best correspond to high level of different components of deliberation.