Wednesday, August 31, 2011

Reaction: Automating the Design of Graphical Presentations of Relational Information

The author starts by describing the challenges involved in building tools that automate the process of designing graphical representation of relational information. The main problems enlisted are codifying graphical design criteria and ways to synthesize wide variety of designs. These are broadly described problems which need to be addressed. Expressiveness and effectiveness are the main criteria by which solutions to these problems can be evaluated. 

The author broadly classifies the existing work in this area into three areas. Content issues, graphic design issues and design variation issues. The authors illustrate the difference between the three areas and discuss examples of work which dealt with the different issues. The paper focuses only on the graphic design issues and the design variation issues. Using an example the author brings out the exponential number of design choices available. These choices are the sample space to choose the final design from. Some more discussion is presented on effectiveness and expressiveness of produced graphical designs. The authors set the focus of the research and the reasoning behind it. They define the broad idea of a graphical language that they are going to use for presenting the research. 

The author then describe how we can generate expressive graphical designs based on the syntax and semantics of the graphical language. They show a few examples of the approach. Then they set on to discuss the effectiveness criteria. Effectiveness doesnt just depend on the properties of the graphic design alone. The features of the perceiver are an important factor in calculating effectiveness. The authors have used Cleveland and McGill's work and extended it to create a ranking mechanism based on the degrees of accuracy. 

At this point the author begins discussing ways to generate the different designs. They start with  naive approach of using all possible designs and then point out that it is not practical in most scenarios. The author then describes the idea of considering graphical presentations as the composition of a set of primitive languages. The authors chooses a list of primitive languages and then defines a set of compositions on these languages using the underlying idea of composing two designs by merging parts that encode the same information. This idea is implemented and some impressive graphical design representations are generated. Much needs to be done to ensure the effectiveness of the graphical representations.