While using data in models other than its original one, it often can not be used as is, some kind of transformations will be needed. Such transformations can be divided into 3 occasions: formal transformation, semantic transformation, and the combine of both. The formal transformation is about the data type, precision, format, etc. All of these formal transformation concerns can be handled by LOD standards. The semantic transformation is about the conceptual and logical aspects of data, and can not be handled by LOD. The semantic transformation even can not be fully done by Ontological methods, because that logical inference can only handle formal semantic problems, most of the real problems can not be formalized completely, and commonsense computation must be introduced to solve them. Today’s commonsense computation is represented by Watson like machine, and still far from mature. Commonsense computation will still become the bottle neck of interoperability in future years.