Robertson, P.

NSP: Natural Scene Paradigm

This text is partially taken from: [ROB91]

For a given data set their is possibly more than one best representation because of the different background of scientists or their differing expectations. The most effective representation depends on the type of information the analyst is interested in and the capabilities of the available representation.

" ... we need a methodology, based on some appropriate information theory of visualization, for choosing data representations to best achieve any specific visualization aim." [ROB91]

One methodology is the Natural Scene Paradigm. It is based on the idea, that the human being is very capable in interpreting natural scenes. There is no problem in finding out connected patterns, distinguishing between fore- and background, recognizing whether scene properties are independent or in relation to each other and so on. So it is very easy to analyze representations based on a natural scene for every scientist independent of his/her special field.

The Natural Scene Paradigm offers an approach to the questions :

1. What mental models most effectively carry various kinds of information ?

   NSP: using clear and easily understood models such as 3D structures or scenes

2.  
3. Which definable and recognizable visual attributes of these models are most useful for conveying specific information either independently or in conjunction with other attributes ?

   NSP: representing data variables by the recognizable properties of the objects or scenes

4.  
5. How can we most effectively induce chosen mental models in the mind of an observer ?

   NSP: inducing them in the observer's mind by using graphics scene simulation techniques

6.  

The practical approach to NSP

1. Extract the structure and the nature of each variable from the data,including
   • the dimensionality and, where known, the parameter relationships; and
   • the type of data (ordinal or nominal, discrete or continuous)
2. Substantiate this information by asking the analyst for clarification of
   • the relationships between parameters and sub-parameters; and
   • the type of each variable
3. Ask the analyst about the important attributes for interpretation, including
   • individual data variables (point, local, global);
   • correlations between them (point, local, global relative to each other attribute for each variable);
   • their relative importance (priority ordering or weighting of these attributes); and
   • any display constraints (for example, analyst requirements forcing particular representations).
4. Match the representations to the interpretation aims:
   • choose (and display) representation optimizing the match, based on strict analyst-specified priorities and constraints;
   • choose (and display) alternative representations optimizing the match on a broader basis;
   • indicate which interpretation aims are satisfied; and
   • allow interactive change of priorities and display regeneration.

Visualization Concepts

Last modified on March 29, 1999, G. Scott Owen, owen@siggraph.org