A Novel Approach for Contextual Clustering and Retrieval of Behavior Trees to Enrich the Behavior of Social Intelligent Agents

Recently, many works have been carried out to find effective ways that can allow for plausibly interactions of social intelligent agents (SIAs) in unpredictable environments a reasonable time. Behavior trees (BTs) knowledge be modeled as graph representation and provide way SIAs effectively interact with the received information. BTs store past experiences then used by adequate human-like when facing new situations (query). One challenge appears agent vast experiences—represented forest BTs—tries retrieve similar BT learn from order plausible current situation cost-effective manner. Cognitive scripts their inherent temporal structure address this where they facilitate use contextual retrieval techniques on represented cognitive scripts. This paper introduces novel hybrid agglomerative hierarchical clustering (H-clustering) similarity-based algorithms: map-and-reduce least common parent (LCP) specific query The model groups BTs, scripts, into compact clusters most one real time without noticeable delay. A comparison was done between performance proposed hybrid-retrieval using semi-structured dataset results showed H-clustering-map-and-reduce is more than H-clustering-LCP it allowed low average 8 × 10−3 s compared 3.1 s, respectively.