My research topic is cross-policy compliance using machine learning. Policy compliance means classifying whether a policy “applies” to a given scenario, “does not apply”, or whether there is simply “not enough information.” The way machine learning models currently solve this problem is to utilize expression trees formed from policies. As the image shows, expression trees have several aspects that a model can utilize to classify a scenario. A model can classify a policy as “not enough information” due to one aspect, and completely discard all other relevant information in a scenario. This greatly limits the applications of the model. My research seeks to allow the model to determine the missing information and generate questions to fill the gaps. The result of my research would be a model that always determines “applies” or “does not apply” which will significantly increase the applications of machine learning for cross-policy compliance.
All rights associated with this image remain with the creator. No use is permitted without their permission. Please contact firstname.lastname@example.org for further information.