tailieunhanh - Characterizing Aesthetic Visualizations

Though Birkhoff’s model does include important components of aesthetics, it does not provide for emotional reactions from interpretation. Birkhoff himself made reference to this, when he pointed out that certain polygons might have associations not accounted for in M that could influence judgement. For example, a cross-shaped polygon may have “positive connotative associations1 .” Later, Berlyne incorporated meaning, as well as complexity and order (or the related property, balance) in the model described next | Characterizing Aesthetic Visualizations Laura G. Tateosian PhD Preliminary Oral Exam Report Knowledge Discovery Lab Department of Computer Science North Carolina State University Raleigh NC 27695-8207 Email lgtateos@ Abstract Visualization scientists would like to engage viewers to encourage exploration. A promising approach to engage viewers is to enhance the aesthetic appeal of the visualization. Psychologists believe that aesthetic judgement can be characterized by a number of emotional and cognitive properties. This project aims to identify some qualities that can be varied in visualizations to influence aesthetic judgment. The properties identified by psychologists provide a good starting point. In this proposal I present three visual qualities related to these properties. I propose to conduct studies in which these three qualities are varied to analyze results statistically and then to seek ways to vary these qualities in a visualization while maintaining perceptual salience. 1 Introduction Visualizations graphical representations of data can provide valuable insights into the datasets they represent. Advances in technology have allowed society to generate and store vast volumes of data. Applications in meteorology genetics networking medical imaging marketing and many other areas rely on collecting and analyzing large datasets. Visualizations can facilitate rapid exploration of such data. The term visualization is also used to refer to the process of converting raw data into images. Often each of the data elements in a large dataset has multiple attributes. For example scientists collecting El Nino data record wind data humidity air temperature sea surface temperature latitude and longitude and other variables at buoys spanning the Pacific Ocean. Thus in the resulting dataset an element represents a buoy and each element has more than six attributes. To visualize an m-dimensional dataset of n elements a set of visual features V V1 . Vm must be