With the described method we created Document Cards for proceedings of the InfoVis 2008 conference. Figure 2.7 shows the document collection as a ma-trix. Furthermore we developed a prototypic application which demonstrates interaction metaphors within each DC. Scalability is shown from a large scale kiosk system to application designs for small handheld devices.
2.3.1 Analyzing the InfoVis 2008 proceedings
Figure 2.7 gives a quick overview of the InfoVis 2008 paper collection.3 Skim-ming over the DCs it is easily perceivable that many graph-based techniques have been accepted to last year’s conference (e.g. DC 0, 4, 7, 11, 13, 17, 22, 24). Such a first impression of the content of the collection is usually based on the images that are depicted. As InfoVis is a visualization conference im-ages are naturally very expressive with respect to what a paper is all about.
3Please note that one of the papers was not automatically parseable and therefore does not show up in the DC matrix.
However, if only the images were given it would be hard to tell in which area of graph-related approaches a paper contributed. The title of the paper and the automatically extracted terms help to clarify this. They reveal for ex-ample that DC 0, 4, 7, and 17 are all papers that deal with graph layout algorithms. DC 13 represents a paper that proposes a visualization approach for large power-law graphs. Omitting unimportant information is important here which is also reflected in the terms of its DC (e.g. “simplification method, edge filter, . . .”). On the other hand the papers of DC 22 and 24 conducted user studies related to graph layout and the impact of data transformation techniques respectively. Consequently, terms such as “experiment, study, hu-man observer, and anova test” appear on their DCs. The above examples show that both the graphical and the textual information of a paper are important to convey its content.
Finally, DC 14 and 15 are interesting because they do not contain any visu-alizations at all (but only schematic diagrams). Although this might seem strange at first for a visualization conference there is a simple explanation:
Both papers contribute with theoretical work in the context of Information Visualization instead of presenting novel visualization techniques.
2.3.2 Interactive tool
We provide an prototypic tool that provides the following interaction features for a Document Card (DC):
• Hovering over the non-image space in a DC shows the extracted abstract of the document as tooltip.
• Hovering over an image displays the image’s caption as tooltip.
• Clicking on a page number (right side of a DC) starts a transition to the full page (see for example DC 3 in Figure 2.7 that has been switched to page mode and shows page 2).
• Clicking on an image starts a transition to the page containing the image.
• Clicking on a term highlights the term in the overview and in all tooltips for this document. Additionally, all images containing this term in their
descriptive text are highlighted. The term density is shown in the page indicator on the right side of a Document Card. The higher the density of the term on a page the less transparent is the corresponding tab. The idea is similar to TileBars [Hea95]. (see DC 12 in Figure 2.7 in which the term “tree diagram” has been selected).
The hovering approaches provide readability of caption and abstract text even when a DC is displayed in small scale. To show connections between images and terms we introduced the idea of highlighting images when clicking on related terms. A video demonstrating the functionality can be found here:
http://vimeo.com/6127783.
2.3.3 Large scale system: The Conference Kiosk
Figure 2.4 shows a system4 which was developed for and exhibited at VisWeek conference 2010. It allows to browse through IEEE Vis and InfoVis publica-tions of 10 years represented as DocumentCards and binned year-wise. The system was exhibit on a touch screen device to lower the entry level for usage and allow fast immersion. The user can navigate, select and collect publica-tions and put them into a private basket. At the end, the citapublica-tions of found interesting publications can be send as email.
To allow high performance for large scale collections, we had to limit inter-action methods when creating the kiosk system. The rendering of Document Cards has been transferred into OpenGL which fastened execution for the cost of programming time.
2.3.4 Application to Small Devices
Document Cards are suitable for smart phone and tablet devices. Figure 2.5 shows examples of both: the mockup of a Document Card on a mobile phone in (a) and the design for a reader application on a tablet device in (b). The screen of the reader application is split into the actual text rendering (middle)
4Hendrik Strobelt, Mathias Heilig, Oliver Deussen: Stay in Touch with InfoVis – Visualizing Document Collections with Document Cards, Interactive Demo / Poster at VisWeek 2010
Figure 2.4: Screenshot of the DC kiosk system for VisWeek publications. While in the browser view (left) the user can pinch and double tap, the menu on the right side allows to modify pan and zoom, gives a detail view for the actual
view while a collection of images and texts display a cloud of previous (left) and following (right) content.
(a) . . . for smartphone
social network graphical user
unified model
Graph Data Network Analysis
Orion Models Graph Data
social network graphical user unified model
(b) . . . for a tablet computer
Figure 2.5: Designs of applications for Document Cards on portable devices.