Master’s Thesis: Visualization Toolbox

A visualization is created following specific rules to encode data in visual features. Visualization Toolbox is an application to discover, explore, and visually analyze datasets from different sources, e.g. publications from digital libraries such as ACM and IEEE, cultural content from services such as Europeana, or the local file system. The Toolbox relies on semantic data models (vocabularies) and their integration (mapping) to automatically generate and configure suitable visualizations. In a nutshell, the role of the Toolbox is to support automated process of providing visualizations for exploring and analyzing data sets from different sources (see Figure 1). 

Figure 1: Toolbox workflow for generating visualizations

Student Tasks
You will be working on three main tasks:

Task 1 – Data Selection and Transformation.  Data attributes are properties that intrinsically describe a piece of data. The first task of this work is the implementation of an interface for retrieving data from different sources and transforming and extracting it’s defining properties. This task involves a proper design so that new data sources can easily be integrated in the future.

Task 2 – Mapping Algorithm. A mapping algorithm encodes data into visual features of a visualization. One way to do so is to map data attributes (extracted in Task 1) to visual variables (visual features of a visualization). This task is about implementing of a well-known algorithm to automatically map data onto visualizations. The algorithm takes the transformed data as input and suggests appropriate visualizations. The implementation should entirely work on the client side, i.e. within the local browser instance. An existing, external visualization recommender is used to rank the suggested visualizations depending on user’s profile.

Part of the task is to devise a technical solution for using OWL/RDF in the browser, which is required to describe the data (RDF Data Cube vocabulary) and the visualization components (Visual Analytics vocabulary). The two ontologies are already available.

Task 3 – Visual Interface. This task involves the implementation of the actual user interface, which provides multiple visualizations for interactive data analysis. The task here is not to develop novel visualizations, but to reuse existing chart implementations (from a library like d3.js) to provide the following functionality:

·         User feedback: Rating and tagging the suggested visualization(s). The feedback contributes to user’s profile, which is passed to the visualization recommender to provide a better ranking of the visualizations. The visualization recommender is already available.

·         Extensibility: Possibility to upload and add new, custom visualization implementations to the toolbox. A visualization implementation consists of JavaScript code and semantic descriptions of the visualization.

·         Interactive analysis: Data exploration is supported through interactions such as brushing and filtering, data aggregation, or showing additional details on demand.