Efficient retrieval of multi-media content depends on the availability of adequate meta-data to indicate the nature of the content. Such meta-data often contains useful hierarchical categorisation but is frequently not consistent between different sources. We outline a method to identify equivalent instances which are described by different meta-data schemata, and to find correspondences between hierarchies, which may be used to improve instance matching. Some successful initial tests of the method on large movie databases are reported.
Trevor P. Martin, Yun Shen