We present A-QuB, a platform that facilitates the exploration, discovery and management of semantic metadata. A-QuB incorporates a multitude of features on top of an intuitive and user friendly environment, in order for both novice and expert users to create and execute complex queries. The platform is agnostic of the underlying conceptual model and con gurable on di erent aspects.
At the other extreme, one can nd custom UIs developed in the context of VREs, such as EPOS, iMarine, Bluebridge, Lifewatch Greece, and others. Despite their capacity to exploit the underlying models, such systems are often too tightly coupled to the domain or schema of interest, hindering portability and restricting the opportunities for collaboration among disciplines beyond the ones originally designed for. It is also not uncommon for such systems to sacri ce expressive features of SPARQL for simplicity in design. Finally, researchers can also rely on traditional SPARQL endpoints provided by widely used triplestores, such as Blazegraph, GraphDB etc, which are usually suited for more expert users.
A-QuB does not only aim at o ering (another) simple query interface for
providing access to the underlying metadata, hiding the complexity of writing
expressive SPARQL queries. Although the look'n'feel resembles similar query
building systems with the goal of reducing the learning curve for the novice user,
the features incorporated are based on a thorough analysis of the requirements of
existing VREs and Research Infrastructures (RIs) (see [
The current A-QuB implementation o ers an experimental metadata catalogue containing metadata from a liated RIs. Imagine a user searching for the projects involving a researcher named Jeremy, whose last name cannot be recalled. The only information is that Jeremy was member of a health organization located around central Europe between 2010 and 2015, and that the metadata for these projects can be found in the EPOS and EKT VREs. Apparently, apart from its simplicity for demonstration purposes, this query is rather vague, calling for some assistance by the system, in order not to overwhelm the user with results.
One key characteristic of A-QuB is that it dynamically executes sub-queries and presents partial results on-the- y, while the user builds her query. The goal is twofold. On the one hand, it helps exploration by presenting results and allowing the user to limit the scope of the query. For instance, once the user nds the Jeremy she is looking for while adding lters, she can select the appropriate instance, without having to re-execute the query from scratch. On the other hand, it helps avoid meaningless queries that return no results. This is achieved by transparently altering the options available during the query building process, eliminating choices that can lead to an empty search space.
A-CuB o ers a combination of di erent ways for searching relevant metadata within the same query. It enables the user to compose queries using keyword search (a person named Jeremy), entity-based search (an entity \Person" with Assistive Query Building for Semantic Data a relation \is member of" on entity \Organization"), time range-queries, if supported by the metadata (the "is member of" relation is restricted between 2010 and 2015), lter-based search (the lter \a project has contact person Jeremy" and the lter \a person is member of an organization unit" are nested conjunctive lters), or geo-spatial queries (the user can interact with a map, in order to select a region in central Europe or give coordinates).
Filter-based search occurs based on an intuitive interaction model (Figure 1(a)) and may involve conjunctive and disjunctive nested queries. As excessive use of nested lters may lead to di cult to comprehend queries, there is an option to limit the depth and/or width of lters.
Geo-spatial queries are constructed with the help of an interactive map (Figure 1(b)), o ering most of the features expected, such as searching by toponyms or geographical regions, selecting speci c instances, etc. Note that in most existing systems, this is a standalone feature, whereas in A-CuB it is incorporated as a simple partial step.
The platform also gives the user the ability to store queries for later use. This is a handy option, enabling the retrieval of complex queries, to be used as templates for constructing competency queries with multiple lters or for making minor adaptations to complex exploration tasks.
Every step performed while constructing the query (i.e. every lter applied, instance selected, search-term used or region set, etc.), is analyzed by the system to construct SPARQL building blocks. These blocks are merged appropriately together, creating the nal QUERY that is executed.
After executing a query, the results can be examined through an internal resolver, allowing simple navigation from any instance of the targeted entity to any of the instances of the related entities. In that way, end users can even continue their discovery after the results are retrieved. For instance, in the running example, the user can explore more information about Jeremy from the returned results, in order to make sure this is the person originally intended.
Finally, the platform provides a number of peripheral functionalities that complement its operation. A multi-factor authentication mechanism is used for granting access to the end users, while Role-Based Access Control ensures that users actions are properly regulated. Importing metadata is also supported, coupling the process with the acquisition of relevant provenance information for facilitating trust and versioning. Also, the platform is con gurable on many different levels, in order to enhance exibility, robustness and simplicity. 2.1
Implementation A-QuB is a modern web platform, implemented in Spring Boot at the back-end, and the MVC AngularJS framework at the front-end. It is executed from the command line as a standalone Maven application, which embeds a Jetty server container (by default) and does not require any prior installed software, which makes it completely portable.
The design of the system is based on the Service-Oriented Architecture. The back-end, which is responsible for the management of Semantic (RDF) Data has
V. Kritsotakis et al. (a) Multiple nested lters (b) Map-based interaction been implemented using a RESTful API on top of the open source version of Virtuoso Universal Server. Any other triple store providing the basic operations (e.g., query, insert etc.) can be used as well; an initial version operated on the Blazegraph triple store. Finally, a lightweight and portable H2 Database Engine is used for storing con guration options and saved queries.
As part of the H2020 VRE4EIC project, A-QuB currently operates using the CERIF2 standard conceptual model, enabling interoperability among diverse research infrastructures. A series of task-based evaluation sessions have been conducted in the course of the project in several events with satisfactory results, as reported in the project's deliverables3
Our future implementation tasks involve the addition of features, such as the ability to export results in various formats and the linking of data with external sources. A more ambitious objective is to seamlessly feed resources retrieved by querying into external tools, e.g., work ow engines, data visualizers etc. 2 https://www.eurocris.org/cerif/main-features-cerif 3 https://www.vre4eic.eu/publications/public-deliverables