The twitter-explorer is an open-source research tool that allows users without programming skills to collect, transform and visualize Twitter data through the paradigm of networks. It represents different types of platform-specific interaction patterns (retweet, mention, quote, reply) as interactively explorable networks, 2D-spatialized through a force-directed layout algorithm. The tool, and more specifically its interactive aspect, which allows users to read tweets of users from within the network visualization, was built out of a necessity to go back to the actual data in order to understand why certain nodes find themselves in the position they are in. We call this process "guided close-reading": the structural overview given by the spatialized network allows the social scientist to find, sort and then analyze relevant content through traditional methods from discourse analysis in order to generate hypotheses about the underlying debate on Twitter.

In this presentation, I would mainly like to discuss the intricacies of tool-building for the social sciences, more specifically the use and interpretability of force-directed layout algorithms, as well as get feedback on the architecture of the tool itself, which is written in Python and JavaScript.