Improving the Scalability of Interactive Visualization Systems for Exploring Threaded Conversations

Abstract

Large threaded conversations, such as those found on YCombinator's HackerNews, are typically presented in a way that shows individual comments clearly, but can obscure larger trends or patterns within the conversational corpus. Previous research has addressed this problem through graphical-overviews and NLP-generated summaries. These efforts have generally assumed a particular (and modest) data size, which limits their utility for large or deeply-nested conversations, and often require non-trivial offline processing time, which makes them impractical for day-to-day usage. We describe here Forum Explorer, a Chrome extension that combines and expands upon prior art through a collection of techniques that enable this type of representation to handle wider ranges of data in real time. Materials for this project are available at https://osf.io/nrhqw/.