Inferring Domain Models to Help Analysts with Belief and Intention Recognition

Inferring Domain Models to Help Analysts with Belief and Intention Recognition Abstract We have shown in past work that belief and intention recognition is a useful tool for identifying potential explanations as to why certain actions are taken in a planning domain. This enables analysts to quickly identify potential future analyses to perform so that they can form a more complete model of the world and the events that have occurred in the past. The issue with belief and intention recognition is that it requires a bespoke model of the domain, which can be difficult for non-experts to build. In this work, we will develop techniques for automatically learning this domain based on natural language summaries of events that have occurred. We utilize develop natural language processing techniques and deep learning to automate the domain modeling process to reduce the authorial burden on those who wish to use belief and intention recognition.