Despite the increasing availability of RDF data, exploring and browsing semantic datasets is still a daunting task for lay users. RDF Surveyor is an easy-to-use visual exploration tool that can be plugged in any CORS-enabled SPARQL 1.1 endpoint without requiring any installation. RDF Surveyor offers an intuitive user interface that gives an overview of the repository contents, supports class navigation, and individual visualisation. RDF Surveyor is very lightweight and can handle large datasets such as DBpedia in a responsive way.
In this demonstration we present RDF Surveyor, a novel and lightweight tool for exploring semantic datasets. It offers an easy-to-use interface that completely hides the RDF and SPARQL syntax from the user, providing an overview of the repository contents, intuitive class navigation, and visualisation of individuals. Notably, RDF Surveyor does not require any installation and can handle large datasets. 2
RDF Surveyor is a Javascript tool available under an Apache 2 license on GitHub.2 A live version of RDF Surveyor is also available at http://tools. sirius-labs.no/rdfsurveyor for testing purposes. The tool runs entirely in a browser without relying on any server component, while target repositories only require a CORS-enabled3 SPARQL 1.1 endpoint.
RDF Surveyor solely relies on the SPARQL endpoint to build the user interface, submitting queries on demand based on the actions performed by the user, and caching results in the session storage of the browser to minimise the number of queries posed. To begin the exploration of a dataset, the user only has to enter the URI of the SPARQL endpoint and optionally the URI of a named graph in a web form. RDF Surveyor then submits a bootstrapping query to get the upper classes in the dataset; when results arrive, further queries are sent to obtain labels, comments, direct subclasses, and number of individuals for the upper classes. All this information is used to create an overview page of the repository with the employed namespaces (Fig. 1(a)) and the list of upper classes sorted by the number of direct subclasses and members (Fig. 1(b)).
The user can then filter out unwanted namespaces, drill down the subclasses of a listed class, or select a class. In the latter case, RDF Surveyor prepares a page with the class URI, label, comment, superclasses, subclasses, and member individuals (Fig. 1(c)) – new queries are sent in case of missing information. The user can navigate the class taxonomy, and there are controls for paginated browsing and searching individuals by label – supported by limit/offset modifiers and regex expressions. If an individual is selected, a page is created with the individual URI, label, comment, class types, datatype properties, and direct and inverse object properties (Fig. 1(d)). RDF Surveyor also tries to find a picture URL and geographic coordinates to feed an image and a map widget, respectively. The user can jump to other individual or class pages by clicking the corresponding links. In every page, there is a search textbox for finding classes and individuals by label in the whole dataset (see the navbar in Fig. 1(a)).
Other features of RDF Surveyor include its RESTful design, so all pages expose consistent URIs that can be safely shared and bookmarked; support for multilingual datasets, taking the browser language as first option and English in other case; and, finally, the user interface is adapted to mobile devices. 2 https://github.com/guiveg/rdfsurveyor 3 https://enable-cors.org/ 3
ID Step description S1 Load DBpedia repository S2 Filter out the owl namespace S3 Expand subclasses of dbo:Work S4 Load the Artwork class S5 Find members of Artwork with breda in the label S6 Load The Surrender of Breda individual
The demonstration will showcase the exploration of DBpedia or any other dataset exposing a SPARQL endpoint with RDF Surveyor. Table 1 presents a sample walkthrough for producing Fig. 1 – the number of queries submitted and the elapsed time are reported, as measured with a 2014 MacBook Pro. Every step took less than 1 second to complete, so the user experience is not compromised.4 Since query results are cached locally, a repetition of the walkthrough does not involve new queries and every step takes then less than 0.1 second.
In Step S1, information about the upper classes is obtained (see Section 2) – note that for every class found, RDF Surveyor asks if it has more than one thousand members, and only requests the actual count in case of a negative answer for performance reasons. S2 is the slowest step due to an expensive query with a STRSTARTS function for detecting subclasses of upper classes beginning with the owl namespace, finding 49 subclasses such as dbo:Place or dbo:Work; subsequent queries compiled information about the classes found.
This work has been funded by the Norwegian Research Council through BIGMED (IKT 259055), HealthInsight (NFR 247784/O70), and SIRIUS (NFR 237898).