Exploring entities in text with descriptive non-photorealistic rendering

We present a novel approach to text visualization called descriptive non-photorealistic rendering which exploits the inherent spatial and abstract dimensions in text documents to integrate 3D non-photorealistic rendering with information visualization. The visualization encodes text data onto 3D models, emphasizing the relative significance of words in the text and the physical, real-world relationships between those words. Analytic exploration is supported through a collection of interactive widgets and direct multitouch interaction with the 3D models. We applied our method to analyze a collection of vehicle complaint reports from the National Highway Traffic Safety Administration (NHTSA), and through a qualitative study, we demonstrate how our system can support tasks such as comparing the reliability of different models, finding interesting facts, and revealing possible causal relations between car parts.