Tag-based algorithms can predict human ratings of which objects a picture shows

Collaborative tagging platforms allow users to describe resources with freely chosen keywords, so called tags. The meaning of a tag as well as the precise relation between a tag and the tagged resource are left open for interpretation to the user. Although human users mostly have a fair chance at interpreting this relation, machines do not. In this paper we study the characteristics of the problem to identify descriptive tags, i.e. tags that relate to visible objects in a picture. We investigate the feasibility of using a tag-based algorithm, i.e. an algorithm that ignores actual picture content, to tackle the problem. Given the theoretical feasibility of a well-performing tag-based algorithm, which we show via an optimal algorithm, we describe the implementation and evaluation of a WordNet-based algorithm as proof-of-concept. These two investigations lead to the conclusion that even relatively simple and fast tag-based algorithms can yet predict human ratings of which objects a picture shows. Finally, we discuss the inherent difficulty both humans and machines have when deciding whether a tag is descriptive or not. Based on a qualitative analysis, we distinguish between definitional disagreement, difference in knowledge, disambiguation and difference in perception as reasons for disagreement between raters.