Research into the generation and utilization of application specific views of reference ontologies. The principle goals of the Ontology Views Project are to:
In developing these methods we will pursue the following specific aims, in collaboration with the National Center for Bioontology (CBio) and the UW computer science department:
The motivation for the ontology views project was born largely out of our experience with constructing and utilizing a large reference ontology, the Foundational Model of Anatomy (FMA). In the case of the FMA, its strongest assets were also its greatest detractors. The FMA is very large and highly detailed. Though it is also very general in nature (e.g. non-application specific) it is rigorous in its adherence to proper modeling principles. These criteria together lend the FMA to many possible applications. However, they have also rendered it cumbersome (i.e. overly large or detailed or principled) for use by any specific application.
For some time potential users of the FMA came to us with similar requests. These requests all had the following basic form, "we really like the FMA, but it is too X for our needs, we really only need Y". Here X is, again, something like "too large" or "too detailed". Y is their description of the FMA derivative that they really desire. These derivatives were generally based on subsets of the whole FMA. The basis for division varied, application to application, but examples include:
Though the desired ontology derivative was generally based on a subset extraction such as those above, it was then often further manipulated to better suit the needs of the application (i.e. classes added, classes removed, properties removed, properties added, etc.).
Initially we handled these requests, for application specific ontology derivatives, in an ad-hoc manner, writing procedural code specific to each new request. As we received more and more requests we realized that we could not continue with this approach. We needed a language for defining the desired application knowledge base (KB) (not always a proper ontology) as well as an engine that could generate the application KB from the definition and the source ontology(ies).
We've modeled our approach to the specification and generation of application specific knowledge sources after the notion of a view in relational databases. In the relational database world, a view is a query expressed in terms of other relations (or views) already found in the DB. The view is conceptually another relation (table) in the DB. If we were to execute the query (e.g. materialize the view) we would indeed have a new table. However, rather than materialize the view, relational database management systems (RDBMS) often store only the non-materialized view and process queries against it by composing incoming queries with the view query to form new queries over the underlying materialized tables. This ensures that the view remains up-to-date with the relations upon which it is based.
Similarly in our project we think of a view as described by a query over some set of input sources. We have specifically targetted semantic web technologies in this research, our input sources are ontologies expressed in any of the languages of the semantic web; RDF, RDFS, and OWL in its various sub-language forms and serializations. A materialized view is restricted only in that it must also be in some form of RDF (closure property). It needn't be a strict subset of the input. In general terms, an ontology view may be any application specific knowledge source specified through query. As the current W3C recommended query language for RDF is SPARQL, we decided to use SPARQL as the basis for our view defintion language.
View all Ontology Views publications here.