Information retrieval and machine learning for probabilistic schema matching

摘要

Schema matching is the problem of finding correspondences (mapping rules, e.g. logical formulae) between heterogeneous schemas. This paper presents a probabilistic framework, called sPLMap, for automatically learning schema mapping rules. Similar to LSD, different techniques, mostly from the IR field, are combined.Our approach, however, is also able to give a probabilistic interpretation of the prediction weights of the candidates, and to select the rule set with highest matching probability.