LinDA - Visualising and Exploring Linked Data

Abstract

The main goal of our work in the context of the LinDA (Linked Data Analytics) project is to o er small and medium sized enterprises (SMEs) possibilities for integrating and consuming data by using Linked Data technologies. One of the major challenges of this project consists in providing user-friendly means of exploring and visualising Linked Data. To achieve this, a Semantic Web application has been created, based on state-of-the-art linked data visualisation approaches, which allows a largely automatic matching and binding of data to visualisations. Hence, in this demo paper we demonstrate the potential of a visualisation framework which is capable of dealing with di erent data formats, serialisations and Semantic Web ontologies.

Keywords

• linked open data
• open data consumption
• visualisation

Introduction

recommendation and pre-con guration and intuitive customization. For this we consider using the visualisation ontology proposed by Voigt et al. besides the other components for conversion, data analysis and (pre-)selection developed within the LinDA project in order to realize and evaluate a generic and usercentered visualisation work ow. 2

LinDA Visualisation Work ow

The visualisation work ow, as depicted in Figure 1, is used for supporting the user in selecting and con guring visualisations and consists of the following steps: 1. Select data: The user starts with the selection of the dataset she intends to visualise (Fig. 2, left). The input data needs to be either in RDF or tabular format in order to proceed to the next step. 2. Select visualisation: Based on the content and format of the data and the semantic descriptions of the available visualisation widgets, a ranking of possible visualisations is computed and presented to the user (Fig. 2, right). 3. Con gure visualisation: After choosing a visualisation from the list of recommended visualisations, the user proceeds to the con guration step. Here, she needs to provide the input necessary for the application in order to map the data to the chosen visualisation (Fig. 3, top). Data conversion may be performed automatically at this stage in case the selected visualisation widget requires a di erent format. 4. Visualise: Finally, following the con guration step, the visualisation and exploration phase of the input data can be performed (Fig. 3, bottom). At this stage, the user can further customise the visualisation or export it in di erent formats, share or publish it or save it for later reuse. In the following sections, we brie y introduce the approaches for the automatic suggestion of suitable visualisations and the automatic mapping of input data to a visualisation being developed.

Recommending Visualisations In order to rank suitable visualisations for RDF input data, a similarity measure between the vocabularies used in the input data and the vocabularies supported by the visualisation widgets is calculated. For RDF input data the list of visualisations can also include visualisations with di erent input formats if corresponding converters are present. If the data is in a non-RDF format, the recommendation is based solely on the format, without similarity calculation or other recommendation strategies.

Mapping Data to Visualisations The approach used for automatically mapping data to visualisations is an extension of the state-of-the-art LDVM [1]. Each visualisation is described by a visualisation model which consists of ( 1 ) an input data type, e.g. cube (table), network or tree, and ( 2 ) con guration options de ning structure and layout of the visualisation. For instance, a bar chart has the cube data model and has the mapping of CSV columns or RDF properties to vertical and horizontal axes as structure options and the axis labels as layout options. The approach for mapping data to visualisations consists of the following steps: First, the appropriate input data type is determined based on the content and format of the data. Then, a con guration form is composed based on the con guration options of the visualisation, which can be pre-con gured automatically if the data is in RDF format (Fig. 3). The content of the con guration form can vary depending on the input data type (e.g. for CSV data, the axes of a bar chart can be mapped to the columns of the input le, while for RDF data, the mapping is realised according to the properties of the instances of a class selected by the user). A pre-con guration is created in both cases, but as in CSV little semantic information is present, it can only be created on a low level. In case of RDF, however, it can be automatically inferred from the RDF

Conclusion and Future Work

In this demo paper, we have introduced a generic approach for automatically suggesting visualisations and binding data of di erent formats and vocabularies, with the goal of providing SMEs with an intuitive way of exploring and visualising data, especially Linked Data. In the future we plan to: ( 1 ) expanding the range of input datatypes, e.g. networks, trees ( 2 ) introduce a generic ontology for describing data and visualisation widgets ( 3 ) improve the user experience by introducing intuitive con guration form templates for selecting and exploring RDF data, and (4) conduct an extensive evaluation through a user study.

References

1. Josep Maria Brunetti , Soren Auer, Roberto Garc a, Jakub Kl mek, and Martin Necasky. Formal linked data visualization model . In Proc. IIWAS , IIWAS ' 13 , pages 309 { 318 , NY , 2013 . ACM.

2. Aba-Sah Dadzie and Matthew Rowe . Approaches to visualising linked data: A survey . Semantic Web , 2 ( 2 ): 89 { 124 , 2011 .

3. Martin Voigt , Stefan Pietschmann, and Klaus Mei ner. A semantics-based, enduser-centered information visualization process for semantic web data . In Semantic Models for Adaptive Interactive Systems , pages 83 { 107 . Springer, 2013 .