Pharaoh: Context-Based Structural Retrieval of Cognitive Scripts

Abstract

Cognitive scripts can act as a basis for representing behavioral tasks and domain knowledge in cognitive systems. Each event in a cognitive script is either temporally or causally linked with preceding and succeeding events. This temporal progression of events is what provides context to a particular cognitive script. In other words, it is this linking that provides a deeper explanation of a key event by defining the settings in which this event occurs (i.e. preceding and succeeding events). Contextual information plays a significant role in the retrieval process of cognitive scripts and needs to be considered in the retrieving process of cognitive scripts from large search spaces. Standard retrieval methods have been used on various unstructured data objects, such as text documents, images, audio, mind maps or videos. Other representations appear in logic-based languages that provide a structure that supports information retrieval based on logical reasoning, such as the Web Ontology Language. However, the application of these methods to structured cognitive scripts is not ideal because of the type of contextual information in cognitive scripts. This article presents Pharaoh, a novel context-based retrieval algorithm for cognitive scripts that can be employed in cognitive systems. Pharaoh relies on semantic structure and keyword-based retrieval to retrieve similar cognitive scripts based on a novel similarity measure between a structured query cognitive script and registered cognitive scripts.