The VisInfo System
Due to technical advances in acquisition, processing and storage of primary research data, increasing amounts of primary research data are available in digital libraries. With the increasing availability of research data new requirements and challenges for precise search and retrieve functionalities are rising. In contrast to text documents, research data, with its graphic vizualisations, places different demands on indexing, searchability and presentation in the information retrieval process. In the project, data analysis processes as well as visual search systems are being studied and developed further, with their prototypical implementation being evaluated in the GetInfo specialized portal by the German National Library of Science and Technology (TIB). The objective is not only the immediate search in research data, but also metadata based searches.
Using a visual cluster algorithm, a huge amount of scientific research datasets of a digital library can be arranged in a self-organizing map and visualized in one Visual Catalog. A example from the project shows daily temperature curves from different BSRN (Baseline Surface Radiation Netowrk) stations all over the world.
Image.: Visual Catalog
For a better understanding numeric data is very often visualized during a scientific analysis. The VisInfo project uses this concept for a content-based search in time-series data and provides a query interface for the definition of a visual query in terms of a curve progression The user can choose a example curve in the Query Editor, upload measured data curves or drawn a curve by hand. Based on predefined similarity notions, similar curves within the search space will be returned and visualized.
Image.: Query Editor
Hits can be display in the Visual Catalog using an additional color mapping. The colormap indicates similar time series patterns with blue colors, unsimilar patterns are denoted with red color values.
Image.: Visual Catalog with color mapping
The Visual Catalog can be visualized with certain metadata filters associated with the corresponding metadata. The white-yellow colormap indicates the number of patterns of each cluster cell that correspond to the filter query (also called density histogram). Yellow colors denote cluster cells with high density.
Image.: Visual Catalog with metadata filter and color mapping