Video Test Collection with Graded Relevance Assessments.

Relevance is a complex, but core, concept within the field of Information Retrieval. In order to allow system comparisons the many factors that influence relevance are often discarded to allow abstraction to a single score relating to relevance. This means that a great wealth of information is often discarded. In this paper we outline the creation of a video test collection with graded relevance assessments, to the best of our knowledge the first example of such a test collection for video retrieval. To directly address the shortcoming above we also gathered behavioural and perceptual data from assessors during the assessment process. All of this information along with judgements are available for download. Our intention is to allow other researchers to supplement the judgements to help create an adaptive test collection which contains supplementary information rather than a completely static collection with binary judgements.