Learning Aesthetics Measure of a Document Page Layout From Designers' Interaction
Ahmadullin, Ildus; Damera-Venkata, Niranjan;
Keyword(s): automated publishing; quantifying aesthetics; designer interaction; probabilistic modeling
Abstract: Aesthetics evaluation of a document layout is typically performed by a designer. Recently there has been proposed a number of automated systems for document creation. We present a new paradigm to automatically quantify document aesthetics, that can be used in the automated document composition techniques. Given a document template, we specify a few aesthetics parameters that can be adjusted to obtain the most aesthetically pleasing layout of a document page. We use graphical interface, where for a given amount of content designers can adjust the parameters to obtain a layout of the highest aesthetic value. The parameters are stored as a feature vector. After obtaining sufficiently large amount of feature vectors, the feature set is modelled by a Gaussian probability distribution. The resulting model can be used in predicting aesthetic value of a new document, or to compose a document of the highest aesthetic value for a given content.
External Posting Date: January 21, 2012 [Fulltext]. Approved for External Publication
Internal Posting Date: January 21, 2012 [Fulltext]