Graphical Query Engine for Knowledgebases

OData support
Supervisor:
Simon Gábor
Department of Automation and Applied Informatics

There are publicly available, cross-domain knowledge bases that contain a vast amount of machine processable, densely interconnected data. We should make this knowledge available to a wide range of people in such a way that they can find answers to their unique questions. Besides conventional, but perhaps less intuitive textual query languages, visual query languages seem to be promising. The Semantic Extension (SemEx), part of a visual modelling system (VMTS) developed in the AUT department utilizes the latter.

However, this tool in itself does not guarantee the formulation of successful queries. For this the user would have be familiar with the concepts and relations used in the knowledge base, but these can differ from their expectations even in everyday topics. As I was contributing to the SemEx project, my goal was to help users build successful queries, regardless of their knowledge about the topics involved or the peculiarities of the knowledge base.

During development I introduced the option to choose concepts based not on keyword search but on citing examples. I have expanded the number of properties assignable to a given entity. I have included additional information to help the user distinguish between similar entries. Furthermore, I introduced the possibility to filter literals, which is a feature that is generally available in query languages.

In my thesis I will briefly explain the basic concepts of the semantic web. I will describe knowledge bases and Wikidata in particular, the semantic dataset queryable by SemEx. I will present VMTS and SemEx in more detail, along with the technologies used during their development, then I will discuss the implementation details, design choices and considerations. I will demonstrate the usability and effectiveness of the new features with case studies for each one of them.

The new features presented here have enhanced user experience at multiple steps of the query building process and contributed to making it easier for users to create more accurate queries with SemEx. These methods can be used in other visual semantic query languages, too.

Downloads

Please sign in to download the files of this thesis.