These results and techniques are meant to be a simple illustration of how some aspects of the general architecture presented can be applied to a real domain. It does not illustrate the use of different PEPs, or global feature calculation. In addition, the selected event search is currently implemented in exactly the same way as the event attribution process. While the accuracy results aren't particularly spectacular, the descriptions generated of the classes are certainly interesting and already showing promise. Most probably as the system is refined and more PEPs are added, then the accuracy (God-willing) will improve.