Crowdsourcing as a Map Reduce Job

MSI proposes to develop automated mechanisms to dissect incoming queries and engage appropriate cloud-based resources to address them. To automate this process, we propose iterative query refinement between an automated analyzer (called the Matchmaker) and human experts. The Matchmaker will use keyword and semantic analysis to identify query topics. These topics will engage people with expertise in those topics who can either take on the subtasks or provide more accurate topic suggestions. When the task results are submitted, an automated Fusion Engine will compare all results with a description of the expected results (submitted by the task creator) and select the best answer. These mechanisms will extend MSI"s foundational system for massively collaborative problem solving that is being developed for a Phase II DARPA SBIR. Our framework, called ePluribus, allows collaborators to explore two key problem solving phases: understanding the situation that brought about the problem and evaluating actions that can modify the outcome of the situation. Similar to the map/reduce model, ePluribus allows people to 1) decompose complex problems into manageable components (map), 2) asynchronously provide solutions to these sub-problems, and then 3) aggregate the proposed solutions to form a collective solution integrating all significant points of view (reduce).