Finding, organizing and analyzing research data (i.e. publications) published in various digital libraries are often tedious tasks. Each digital library deploys their own meta-model and technology to query and analyze the knowledge (in further text, scientific facts) contained in research publications. The goal of the EU-funded research project CODE is to provide methods for federated querying and analysis of such data. To achieve this, the CODE project offers a platform, that extracts scientific facts from research data and integrates them within the Linked Data Cloud using a common vocabulary (i.e. meta-model). To support users in analyzing scientific facts, the project provides means for easy-to-use visual analysis. In this paper, we present the web-based CODE Visualization Wizard, which aims to analyze research data visually with an emphasis on automating the visualization process. The main focus of the paper lies on a mapping strategy, which integrates various vocabularies to facilitate the automated visualization process.
Digital libraries, which control the lifecycle of research publications (i.e. publishing and making them accessible for certain communities) mainly expose the research knowledge using domain-specific meta-models and technologies. Moreover, they only focus on some structural attributes and often don’t consider the content of the publications. This domain-specificity and weakness in specifying querying attributes limit the ability to effectively find desired information, since the number of published content is continuously growing. The goal of the CODE1 [4] [5] project is to offer a solution for this issue by providing a platform that structures (heterogeneous) research data using the RDF Data Cube Vocabulary2 and releases them as Linked Data.
The RDF Data Cube Vocabulary is a generic vocabulary used to describe quantitative data (e.g. research results from tables). To simplify the analysis of this data, the web-based CODE Visualization Wizard3 has been developed, which integrates several visualizations. To achieve a Linked Data-based visualization, these visualizations (e.g. charts) should also be described semantically. For this purpose, we defined the Visual Analytics (VA) Vocabulary4 in the form of an OWL ontology. This vocabulary is an interface between the RDF Data Cube and visualization-specific technologies, and together with the RDF Data Cube Vocabulary it forms the basis for automating the visualization process.
In this paper, we summarize the current status of the CODE Visualization Wizard and its ongoing research. 2
Semantic description of visualizations using RDF is a new research topic and the literature, up to now, offers just a few related publications. The most significant research, the Statistical Graph Ontology [3], comes from the biomedical domain and presents a new approach to annotate visualizations semantically.
While the Statistical Graph Ontology provides a sophisticated ground for describing statistical graphs, some key issues (e.g. the description of size and color as visualization component or the datatype of a visualization component etc.) for our applications were missing. This is why we have extended this vocabulary for our Visualization Wizard.
At Stanford University, an interactive Web-based visualization system, the Vispedia [1], has been developed to visualize heterogeneous datasets. The visualization process of Vispedia is based on the integration of the selected data into an iterative and interactive data exploration and analysis process enabling non-experts to more effectively visualize the semi-structured data available. Vispedia was an inspiration for the Visualization Wizard, but being a Wikipedia plugin it only supports visualization of Wikipedia data. Also, it does not provide automatic binding of heterogeneous data onto visualizations. 3
Approach for Automated Visualization Support
In contrast to other available solutions for visualizing Linked Data [2], the CODE
Visualization Wizard automatically suggests suitable visualizations based on
(
Our VA Vocabulary is used to represent the information about visualizations. It describes the visualization axes and other visual channels, such as color or size of visual symbols, used to visually represent the data. The vocabulary also describes suitable datatypes that can be represented by the axes and visual channels, including the allowed occurrence of the axes and visual channels. The definition of the occurrence is important to identify whether the axes or the visual channel can be instantiated only once (e.g. bar chart x-axis) or multiple times (e.g. parallel coordinates x-axis). In fact, this model is technology-independent and used by the Visualization Wizard to generate the specific visualization code. We use in our Wizard the D36 visualization library and Google Charts7 to create our visualizations but as mentioned above, it is possible to use other technologies.
Currently, the Visualization Wizard supports nine different charts and a table. For the integration of each new visualization, a generator needs to be implemented, which has well-defined interfaces and can be plugged-in to the Visualization Wizard easily.
Mapping Vocabularies: The mapping between both mentioned vocabularies, the RDF Data Cube and the VA Vocabulary, is a relation from dimensions and measures of the RDF Data Cube (i.e. cube components) to the corresponding axes and visual channels of the visualization. The mapping combinations will be found based on the structural compatibility and on the datatype compatibility between a RDF Data Cube and visualizations.
The number of the dimension and measures in a RDF Data Cube is
unbounded. The possible combinations (i.e. in the format dimension: measure) for
each RDF Data Cube are: (
The challenge of the first iteration in developing the CODE Visualization Wizard was to show that pitfalls of traditional visualization principles, such as the need for the manual work and high maintenance while visualizing datasets, can be effectively overcome by describing data and visualizations in dedicated vocabularies and by mapping these vocabularies. From the technical viewpoint, the main challenge was to automatically determine the right mapping between instances of the RDF Data Cube and the existing visualizations. Another, and
G N I P P A M S E I R A LBU RDF Data Cube Vocabulary A C O V
E
C
O
R
IILZTPANO Dat1aset
A
U
S
I
V (
Visual Analytics Vocabulary Visual Analytics Vocabulary Visualization Technology
x-Axis y-Axis more serious challenge was to determine only valid suggestions among the provided visualizations.
The ongoing topics, which are parts of the project’s next iterations, are (
The development of the prototype will continue throughout the rest of the year, leading to a final evaluation at the beginning of 2014.
Acknowledgement This work is being developed at the Know Center within the CODE project funded by the EU Seventh Framework Programme, grant agreement number 296150. The Know-Center is funded within the Austrian COMET ProgramCompetence Centers for Excellent Technologiesunder the auspices of the Austrian Federal Ministry of Transport, Innovation and Technology, the Austrian Federal Ministry of Economy, Family and Youth and by the State of Styria. COMET is managed by the Austrian Research Promotion Agency (FFG).