Abstract
Combining different strands of media into a multi-media stream can assist as a means of improving communication. Providing text and images together can provide a better means of communicating than providing either medium in isolation. Indexing and retrieving multi-media information in a timely and efficient manner has become an area of active research, as real time retrieval of such data has become possible through advancing technology. Where related image and text material are stored together, the text is an additional source for information on a related image, information which may be used for classification purposes, particularly if the text has been produced by an individual who is an expert in the domain to which the image relates. Fine art is a good example where the co-ordination of art and writing about art can enhance communication. Artists create complex visual images and art critics explain, elaborate and classify them. A study of the texts associated with fine art images provides a vehicle for investigating the links between images and text and also a means of investigating how the information contained within these texts may be used to aid in image classification. Further, the domain of fine art is constrained, which will assist in detecting common themes that may exist in the texts related to fine art images. In this thesis we present the results of an examination of art critics’ writings, where these results are used to assist in classifying paintings on the basis of their semantic properties. We look at the classifications used by critics, and the ways in which they critique a painting. Our results provide evidence that non-experts are able to understand the terminology of art critics, even when these non-experts are unfamiliar with this terminology. Further we provide evidence of recognisable structure in art critics’ writing. This structure is used as a basis for the development of a system archetype, Virtual Gallery Manager (VGM), which embodies an approach to automatically indexing and retrieving fine art paintings, through the identification of relevant terms in the associated critical text.