A Preliminary Approach to Referring to Groups of Objects in Images

In order to talk about an image it is of special interest to be able to refer to the groups of objects that appear in it. To do that, we have to form noun phrases in which properties are combined to discriminate the target group from the rest. These phrases are part of what is known as plural referring expressions in the NLG field. The construction of this type of expressions is a complex problem, given the graduality of both the visual properties that can be used to make them up and the concept of group itself, among other reasons. In this work we propose the use of distance as the inducing element of the notion of group of objects in the image and we also show how results from the theory of fuzzy sets and the related theory of representation by levels (RL) can be used to assess the referential success of referring expressions taking into account the aforementioned graduality of the involved concepts.