A Distributed Approach to Coreference Resolution in Multiagent Text Analysis for Ontology Population

In the paper we describe a multi-agent ontology-based approach to coreference resolution and information extraction from input data. We define special class agents which correspond to ontology classes. They analyze current information in the corresponding retrieved instances. Results of this analysis are used for evaluation of instances’ attributes, for detection of instances duplicates and equivalents, for fixing coreferential relations, and for assignment of worth of information connections used in disambiguation. The class agents act in the framework of our multi-agent approach to semantic text analysis for ontology population.