Flashgroups: Automatically Forming Collaborative Groups of Analysts around Prioritized Information

Analysts are faced with information overload in two major arenas: a) an overwhelming amount of information, where only a small subset may be relevant to their task, and b) a huge number of other analysts, a few of whom may be investigating related subjects and able to provide key insights. Here, we propose Flashgroups, a novel tool that helps an analyst discover key groups of information and other analysts who can help achieve her task. As the analyst interacts with the proposed tool, a real-time machine learning algorithm learns a detailed user model that captures the analyst's current intention. Using this model, our proposed system can recommend highly relevant groups of information for the analyst's underlying task. When multiple analysts are using our tool, the machine learning algorithm can transfer discovered knowledge between analysts, and connect the ones who have aligned goals in their current tasks to form collaborative teams, or flashgroups, around specific topics. In Phase I, we will evaluate flashgroups around a database with tens of thousands of documents, and tens of users, thus demonstrating the viability of our proposed strategy for tackling the formation of context-specific collaborative groups of analysts.