Professor Stephen Brown
Although textual resources comprise the single most comprehensive record of British photographic exhibitions in the nineteenth and early twentieth centuries, they are not sufficient where research questions require interrogation of the images to which they refer, such as “What kind of compositional techniques did particular photographers use? How did photographers record a particular genre such as museum displays? How did photography support scientific enquiry and how were the techniques for producing scientific images refined over time?” Exhibition catalogue entries need to be related to the photographs but early exhibition catalogues were often devoid of pictures, relying instead on written descriptions of the image. Thus assigning a specific exhibition catalogue reference to a specific image can be a complex and involved process. Image collections are increasingly being published online, search engines are becoming increasingly powerful, creating a timely opportunity to match photographs with their original exhibition catalogue entries without travel to numerous archives. Amongst the visual arts, photography is unique in that multiple versions of the same image can be produced and exhibited simultaneously at diverse locations. Photographs were commonly exhibited/published more than once, at different times, with different titles and even by different people, making precise associations between exhibit records and images difficult. This project is developing and testing computer based “finding aids” that will be able to recommend potential matches between historical exhibition catalogue entries and images of photographs in online collections even where there is not a precise match. Possibilities include allowing researchers to:
Establish links between photographs, photographic technology and associated apparatus by incorporating images of equipment &c., for the first time.
Reassemble the visual elements of an exhibition and through this more fully understand its aesthetic, social, cultural context.
Relate the appearance of an image to processes associated with its making.
This project is investigating the potential of combining Probabilistic Record Linkage with fuzzy clustering to identify co-reference between exhibit records and published images based on attribute similarity.