Automatic Semantic Transformation between Geo-Ontologies

整个项目旨在开发地理空间数据集语义整合技术。为了做到这一点，必须定义不同领域本体之间的对应关系。目前，这通常是手动完成的，因为在没有先验知识的情况下，如果不是不可能的话，也很难识别对应关系。在这个建议中，我们要调查和利用几何线索（加上拓扑和语义的）来推断语义对应。提出了以下假设：当对象描述描述空间上相似的对象（即在相同空间位置的对象）时，或者如果它们具有相似的几何特征，则对象描述涉及相同的现象。在第一阶段，重点放在同一位置的数据与类似的几何和语义分辨率的整合。第二阶段的研究将从两个方面放宽这些限制：1）整合不同位置的数据集，2）整合不同分辨率的数据集。为了获得必要的可比对象实例，在第一种情况下，将确定基于几何和拓扑特征的分类规则。在第二种情况下，将应用泛化规则。作为一种方法来获得的对应关系-特别是层次对应关系-我们将使用数据挖掘的关联规则。

The overall project aims at developing techniques for the semantic integration of geospatial data sets. In order to do so, correspondences between different domain ontologies have to be defined. Currently, this is typically done manually, as without prior knowledge it is difficult if not impossible to identify correspondences. In this proposal we want to investigate and exploit geometric cues (plus topologic and semantic ones) to infer semantic correspondences. The following hypothesis is made: object descriptions relate to the same phenomenon when they either describe spatially similar objects (i.e. objects at the same spatial location), or if they have similar geometric characteristics. In the first phase the focus was laid on the integration of data of the same location with similar geometric and semantic resolution. The research in the second phase will relax these restrictions in two ways: 1) integration of data sets of different location, 2) integration of data sets of different resolution. In order to derive the necessary comparable object instances, in the first case classification rules based on geometric and topologic characteristics will be determined. In the second case, generalization rules will be applied. As a method to derive the correspondences – especially the hierarchical correspondences – we will use association rules from Data Mining.