|Title:||A New Framework for Querying Semantic Networks|
|Authors:||Katerina Tzompanaki, Martin Doerr|
|Publication:||MW2012: Museums and the Web 2012|
Repositories of textual content can successfully be accessed by text search engines using combinations of keywords. "Open World" semantic networks of RDF triples however, the core of the Semantic Web and the state-of-the-art of metadata repositories for museum and other cultural information, are not easily accessible by such a global query paradigm. Only a small part of the information can be related to words in labels. Querying individually dozens of different kinds of properties as in ISO21127 (CIDOC CRM), many of which are optional, leaves a huge recall gap compared to text retrieval. On the other hand a global restriction to "core metadata", to a flat set of very few properties as in Dublin Core or VRA, deprives the systems of the reasoning capability about the integrated knowledge in the semantic network. We propose and have been implementing a new query paradigm: A very small set of "fundamental" categories and relationships, as we are used to from core metadata, are presented to the user in the form of a system of invisible to them complex deductions from a rich underlying network of more specialized actual metadata, rather than being primary documentation elements. Using as schema ISO21127 and specializations (extensions) of it, we are practically able to achieve high recall rates (even 100%) with a compact set of very simple, comprehensive questions in an application comprising general cultural heritage data about museum and archaeological objects and digital provenance data of 3D digitization of the latter. But except from efficiency we also provide simplicity, as the user does not need to be experienced in any way with the concepts of the CIDOC CRM in order to obtain the desired results. Application of the presented framework can easily be adjusted to many other domain ontologies and user preferences.