• Home
• About
• Contact Us
• Activities
  • By Discipline
  • By Method
  • Core Activities
  • Distributed Activities
• Resources
• Publications
• Get Involved
• e-Science

Annotating Image Archives To Support Literary Research Seminar Report

Report by Omer Rana, Paul Rosin, David Skilton, Julia Thomas

Introduction

This report summarizes the discussion at the AHRC ICT Methods Network funded workshop on ‘Annotating Image Archives to Support Literary Research’, which took place at the Welsh eScience Centre (Cardiff) on 10 and 11 October, 2007 (in collaboration with the Centre for Editorial and Intertextual Research).

The Centre for Intertextual Research (CEIR) at Cardiff has completed an AHRC funded project to create a Database of Mid Victorian Wood Engraved Illustrations (DMVI). Taking 1862 as a representative sample year, it contains 868 Victorian literary illustrations, drawing in the main on two major collections: the periodical illustrations of the 1860s and 1870s in the School of Art Museum and Gallery, Aberystwyth University, and the Forrest Reid collection in the Ashmolean Museum, Oxford. Each image in the database is accompanied by iconographic annotations, enabling searches to be performed not only on the title, artist, author, and other material and bibliographic properties, but also by the pictorial content of the illustrations themselves. Successful iconographic description requires a standardized vocabulary and a set of keywords (divided into a set of predefined “facets”), which allow the database user to consistently locate illustrations matching their search terms. Identifying the best representation scheme to provide these annotations has been the basis of discussion between researchers within CEIR, the Welsh eScience Centre and the School of Computer Science.

Based on research within computer science, for instance, such keywords and their relationships may be expressed using Resource Description Framework (RDF) and the Ontology Web Language (OWL). In addition, there could also be potential benefit in adopting techniques from Content Based Image Retrieval (CBIR) research – which has primarily focused on using the content of an image to match against similar types of images within a database. In this approach, the focus is on comparing a template image with those in an archive, rather than providing explicit annotations/keywords that may be associated with the image.

Read the report...