Probabilistic model for schema understanding and matching

Schema matching is the problem of finding a correspondence between schemas: relational database schemas or Web repositories. The ad-hoc approaches perform poorly on real-world schemas due to the abundance of non-dictionary abbreviations. Our approach to schema matching is based on deep understanding of the schema, using an integrative probabilistic generative model as a framework for comparing candidate interpretations. In empirical evaluation, the proposed model performed well in interpreting relational schemas and exhibited high resistance to noise in the model parameters