Talking about quantities in space: Vague quantifiers, context and similarity

Abstract

In this paper we examine how vague quantifiers, such as few, several, lots of, map onto non-linguistic number systems. In particular our focus is to examine how judgements about vague quantifiers are affected by the presence of objects in visual scenes other than those being referred to. An experiment is presented that manipulated the number of objects in a visual scene (men playing golf; the ‘focus’ objects) together with the number of other objects in those scenes and their similarity — in terms of form (women or crocodiles) and function (playing golf, not playing golf) — to the focus objects. We show that the number of other objects in a scene impacts upon quantifiers judgements even when those objects are in a different category to the focus objects. We discuss the results in terms of the mapping between the large approximate number (estimation) system and language

Learning Context on a Humanoid Robot using Incremental Latent Dirichlet Allocation

“…It is a matter of consensus across fields that context processing is an essential part of embodied cognition (e.g., psychology [2], language [4]- [6], AI [7], robotics [8], [9] and computer vision [10], [11]). Schank and Abelson [7] argued that reasoning about situations in daily life relies on "scripts" that inform reasoners about the prototypical features of these situations.…”

“…Barsalou, for example, has advocated the necessity for concepts to be situated [2], [15]; in other words, for an abstract concept to be related to concrete contexts. Coventry et al [6] studied the difference between geometric and functional contexts in the use of spatial prepositions ("over" vs. "above") and of linguistic quantifiers ("few" vs. "many" vs. "several").…”

“…The loop of (a) context deduction, (b) probabilistic update of concept web, and (c) reiteration of MRF, as described in Section IV-D and Equation 6, also visualized in Figure 6, is utilized for refining predictions in context. Combined results for all 15 test objects are demonstrated in Figure 13, which also show an improvement of performance for the context-guided recognition.…”

Learning and using context on a humanoid robot using latent dirichlet allocation

“…It is a matter of consensus across fields that context processing is an essential part of embodied cognition (e.g., psychology [1], language [4]- [6], AI [7], robotics [8], [9] and computer vision [10]. Shank and Abelson [7], for example, argued that reasoning about situations in daily life relies on "scripts" that inform reasoners about the prototypical features of these situations.…”

“…Barsalou, for example, has advocated the necessity for concepts to be situated [1], [14]; in other words, for an abstract concept to be related to concrete contexts. Coventry et al [6] studied the difference between geometric and functional contexts in the use of spatial prepositions ("over" vs. "above") and of linguistic quantifiers ("few" vs. "many" vs. "several").…”