User-defined relevance criteria for image retrieval : an exploratory study

Abstract

Visual information, communicated through images is utilized in many fields including medicine, journalism, education and everyday life (Choi & Rasmussen, 2002:696). The advent of digitisation has seen an explosion in the volume of images available to people. This proliferation of images has led researchers to explore how best to provide intellectual access to visual information. Research of this type aims at producing automated systems that offer the best information retrieval processes for people who require images to meet their information needs. The research study has analysed how people judge the relevance of an image retrieved by an automated system to their information need. The content of participants' responses was analysed to produce a set of relevance judgment criteria specific to visual information. These criteria were compared with criteria developed in a text-based study. From this comparison it was possible to assess the extent to which text-based relevance judgment models can be applied to visual information needs. The results of the content analysis were then used to assess the extent to which participants' relevance judgment changed when selecting images to match queries posed at different levels of meaning.