Semantic product properties and shapes - what your customer already knows

One of the problems with respect to the integration of 3D shape models into more sophisticated applications is their restriction to the pure visual features of an object. Current research intends to find means to integrate other aspects, such as structural and semantic properties, enabling a more comprehensive and more powerful representation. Most approaches towards semantic information and products are data-driven, analysing either existing vocabularies or atomic geometrical features. Instead, we have looked at human perception, and have done a user study on spoken and drawn utterances. The results indicate an additional class of non-visual shapes, which we suggest to name 'perceptional shapes'. These shapes cannot be derived from inference on the given shapes or taxonomies. This paper introduces a method and example of how such information can be obtained and structured, and briefly discusses how it may be deployed in future applications.