3 Integrating Key Concept Extraction in Cupboard and Watson

Besides providing better support for exploring ontologies within the NeOn Toolkit, an important application of key concept extraction is to offer concise, easy to understand summaries of potentially big ontologies. Such summaries are particularly useful in systems where the user needs to obtain a quick overview of the content of multiple ontologies, for example for the purpose of ontology selection. For this reason, we developed a service to generate small, concise snapshots of key concept-based visual summaries of ontologies, to be used in the Cupboard and Watson systems.

Figure 14. Visual summary of the SWRC ontology, as shown in an ontology space in Cupboard.

In principle, this service is based on ideas similar to the ones implemented by KC-Viz. However, in contrast with KC-Viz, it is limited to providing small, static images that capture what is intended to be the “core” of a given ontology.

The service takes as input the URI of an ontology which has to be present either in Watson or Cupboard, and produces a small image displaying the names of the top six key concepts in this ontology - see Figure 14. The images also display links between the concepts, to represent the fact that they are connected. These links do not represent any particular relation, but give instead an indication of existing relations or paths that connect the two concepts in question. In addition, the size of the label associated to each concept denotes the importance of the concept according to the key concept extraction mechanism.

Technically, this service is implemented as a simple REST API, which takes as parameter the URI of the ontology, and redirects the client to a png image, which is either generated on the fly, or stored as a result of a previous invocation.

Thanks to this service, visual summaries are provided for ontologies included in ontology spaces in the Cupboard system. As an illustration, figure 15 shows part of the ontology space associated with the ontologies for the Semantic Nomenclature case study in NeOn. Hence, the user can get a quick overview of the content of a collection of ontologies, which can be useful to quickly identify a particular ontology in a potentially large collection. The service has also been used in the web interface of Watson, to show a summary of an ontology’s content in the page showing the key metadata of the ontology.

Figure 15. Part of the front end of an ontology space in Cupboard.

4 Conclusions

This report describes the new functionalities implemented in the latest version of KC-Viz, an innovative tool for visualizing and navigating ontologies, which uses the Key Concept Extraction algorithm originally developed in the OpenKnowledge project5 to provide abstraction mechanisms based on dynamically generated summaries of an entire ontology, a network of ontologies, or the neighbourhood of a class.

Our next step will be then to perform some rigorous empirical studies, to validate our hypothesis that such abstraction mechanisms make it easier for users to make sense of the content of an ontology, especially in scenarios where the ontology is unfamiliar and users need to gain an initial understanding of the scope of the ontology.





5 http://www.openk.org.