Eswc2009

The document proposes ReduCE (Reduced Coulomb Energy), a machine learning method for approximate classification of individuals using knowledge bases. ReduCE trains a model called an RCE network on a set of pre-classified individuals and background knowledge from the knowledge base. The trained RCE network can then be used to classify other individuals by calculating their similarity to the network patterns. Experiments show ReduCE achieves high matching rates and low error rates compared to a standard reasoner. Future work will explore extensions like forcing binary responses and adding probabilities to assertions.