Flexible or strict taxonomic organization ? Impact on culturally-specific knowledge transfer

This work compares methods for constructing feature-based ontologies that are supposed to be used for culturally-specific knowledge transfer. The methods to be compared are the Terminological Ontology (TO) [1], a method of constructing ontology based on strict principles and rules, and the Infinite Relational Model (IRM) [2], a novel unsupervised machine learning method that learns multi-dimensional relations among concepts and features from loosely structured datasets. These methods are combined with a novel cognitive model, the Bayesian Model of Generalization (BMG) [3] that maps culturally-specific concepts existing in two cultures in an effective and intuitive manner.