From Implicit Patterns to Explicit Templates

Claims for Topic Maps interoperability have been standing so far on formal representation of subjects by topics, and identification of topics by controlled vocabularies or Published Subjects. Those features allow topics to be merged on the basis of common identifiers, leading to aggregation of relevant information or knowledge expressed by characteristics of the merged topics. But what should happen to associations at merging time remains an open issue in many cases. According to current standard processing requirements, only exactly identical associations will be merged. Default standard definition of association templates, there is so far no standard rules for discovering and dealing with associations having similar patterns of roles, or representing the same association type with slightly different patterns. And default proper association merging rules, Topic Maps are at risk to be cluttered up with redundant or non-consistent associations, after several successive merging. A review of how current Topic Map applications deal with those questions is presented. It's shown that they generally use explicit association templates mechanism to identify, create or check patterns. Although there is no official standard, proprietary solutions are quite similar, and the road to adoption of standard simple rules could be fairly easy. An extension of XTM syntax to express templates, conformant to current practices, is proposed. An use case is presented, where a Topic Map repository in an integrated information system is updated using a text mining tool, a controlled vocabulary and a workflow of semi-structured documents. Similar, redundant or overlapping associations are likely to be generated, and the update process has to deal with them Further applications are considered, like checking for template consistency before merging Topic Maps, extracting templates from existing Topic Maps, or developing libraries for specific application domains, extending the notion of Published Subjects to association templates.