Abstract |
Today, numerous Knowledge Graphs, expressed in RDF, play a crucial role in consolidating
and integrating data from diverse sources. It would be very valuable to delve into the
analysis of these graphs for enhanced insights and understanding. However, formulating
analytical queries over Knowledge Graphs in RDF is a challenging task due to the complexity
and scale of these graphs that presupposes familiarity with the syntax of the corresponding
query language (i.e. SPARQL) and the contents of the graph. To alleviate this
problem, we introduce an interactive model to assist users in formulating analytic queries
over complex RDF Knowledge Graphs, irrespective of their schema structure. This is particularly
crucial, since in non-star-schema-based knowledge graphs, the presence of nonstar-
schema relationships requires a more complex querying approach. To provide an intuitive
interface,we leverage users’ familiarity with Faceted Search systems, andwe extend
it for enabling the formulation of analytic queries in a user-friendly way. In particular, we
start from a general model for Faceted Search over RDF data, and we extend it with actions
that empower users to formulate simple and complex analytic queries, as well. These actions
correspond to queries of a high-level query language for analytics, named HIFUN,
that we then translate to SPARQL.Most, the proposed interactive model serves a dual purpose,
addressing not only the formulation of analytic queries, but also the formulation of
exploratory queries; it lets users transition seamlessly from locating resources in a Faceted
Search manner to performing in-depth analyses of the underlying RDF Knowledge Graph.
This accommodates the diverse needs of users, enabling both flexible and dynamic exploration
and analysis of the graph. Additionally, the formulation of queries, including nested
ones, is gradual acknowledging the iterative nature of data analysis. This process involves
repeated and refining steps, allowing users to deepen their queries as they gain insights
into the graph’s structure and content. Overall, the main contributions of this dissertation
are: (i) we present a user-friendly interface for intuitively analyzing RDF Knowledge
Graphs, and (ii) we formally define the state-space of the interaction model as well as the
algorithms needed to produce user interface actions. We also describe and provide a complete
implementation of the model and the relating algorithms, showcasing its feasibility
in real-world scenarios. This emphasizes the practical applicability of our approach, making
it valuable both for analysts and ordinary users dealing with RDF Knowledge Graphs.
Finally, we discuss the results of a user evaluation, providing evidence of the method’s acceptance.
This empirical validation not only underscores the effectiveness of our model,
but also sheds light on future development. In essence, our research not only tackles the complexities of formulating analytic queries over RDF Knowledge Graphs, but also emphasizes
the friendliness and acceptance by users.
|