Anne Laurent, full professor at the University of Montpellier, France

"Flexible query answering systems and data lakes"

Abstract: Data lakes have emerged these past few years as a novel paradigm to deal with very large volumes of data from various and numerous data sources and non-predetermined use. No formal definition has yet been proposed, although they are more and more implemented in companies. They are distinguished from existing decisional information systems based on some features that will be discussed in this talk. We then propose to explore what flexible query systems could do for such systems and how they might need to be extended.

Jens Lehmann, full professor at the University of Bonn, Germany

"Question Answering over Knowledge Graphs: State of the Art and Future Perspectives"

Abstract: Being able to access knowledge graphs in an intuitive way has been an active area of research over the past years. In particular, several question answering (QA) approaches which allow to query knowledge graphs in natural language have been developed as they allow end users to access knowledge without needing to learn the schema of a knowledge base and learn a formal query language. The talk will first describe the historical development of question answering systems and the challenges that arise in them. This is followed up by a presentation of the major research directions in the field (covering both classical natural language processing and "end-to-end" trained approaches) along with current approaches for solving them. Finally, open problems and the future perspectives of the field are discussed.

Alexandra Poulovassilis, full professor at the Birkbeck University of London, UK

"Supporting Users’ Flexible Querying of Knowledge Graphs"

AbstractSemantic web, information integration and information extraction technologies are enabling the creation of vast information and knowledge repositories, for example in the form of RDF Linked Data. Although applications are increasingly taking advantage of such knowledge graphs in domains such as web information retrieval, formal and informal learning, health informatics, entertainment and cultural heritage preservation, there is a need for mechanisms to support both application developers and end-users in discovering useful information in large, complex, heterogeneous graph datasets. 

Users may not be familiar with the full structure and content of such datasets and may need to be assisted by querying systems that do not require queries to match exactly the data being queried, but rather can automatically make changes to the query so as to help the user find the information being sought. Examples of possible approaches are: approximate querying that does not require users’ queries to match exactly the graph structure but automatically makes changes to the query so as to help the user find relevant paths through the data; query relaxation, based on domain knowledge relating to the data; and similarity-based matching.

This talk will discuss ongoing work on flexible querying of knowledge graphs using SPARQL, focusing on proposed language extensions to SPARQL 1.1, their semantics and complexity, algorithms for query evaluation that incrementally return query answers at increasing ‘distance’ from the user's original query, and directions of future work.