On Visually-Grounded Reference Production: Testing the Effects of Perceptual Grouping and 2D/3D Presentation Mode

Rethinking symbolic and visual context in Referring Expression Generation

Situational context is crucial for linguistic reference to visible objects, since the same description can refer unambiguously to an object in one context but be ambiguous or misleading in others. This also applies to Referring Expression Generation (REG), where the production of identifying descriptions is always dependent on a given context. Research in REG has long represented visual domains throughsymbolicinformation about objects and their properties, to determine identifying sets of target features during content determination. In recent years, research invisual REGhas turned to neural modeling and recasted the REG task as an inherently multimodal problem, looking at more natural settings such as generating descriptions for objects in photographs. Characterizing the precise ways in which context influences generation is challenging in both paradigms, as context is notoriously lacking precise definitions and categorization. In multimodal settings, however, these problems are further exacerbated by the increased complexity and low-level representation of perceptual inputs. The main goal of this article is to provide a systematic review of the types and functions of visual context across various approaches to REG so far and to argue for integrating and extending different perspectives on visual context that currently co-exist in research on REG. By analyzing the ways in which symbolic REG integrates context in rule-based approaches, we derive a set of categories of contextual integration, including the distinction betweenpositiveandnegative semantic forcesexerted by context during reference generation. Using this as a framework, we show that so far existing work in visual REG has considered only some of the ways in which visual context can facilitate end-to-end reference generation. Connecting with preceding research in related areas, as possible directions for future research, we highlight some additional ways in which contextual integration can be incorporated into REG and other multimodal generation tasks.

Contrast perception as a visual heuristic in the formulation of referential expressions

We hypothesize that contrast perception works as a visual heuristic, such that when speakers perceive a significant degree of contrast in a visual context, they tend to produce the corresponding adjective to describe a referent. The contrast perception heuristic supports efficient audience design, allowing speakers to produce referential expressions with minimum expenditure of cognitive resources, while facilitating the listener's visual search for the referent. We tested the perceptual contrast hypothesis in three language-production experiments. Experiment 1 revealed that speakers overspecify color adjectives in polychrome displays, whereas in monochrome displays they overspecified other properties that were contrastive. Further support for the contrast perception hypothesis comes from a re-analysis of previous work, which confirmed that color contrast elicits color overspecification when detected in a given display, but not when detected across monochrome trials. Experiment 2 revealed that even atypical colors (which are often overspecified) are only mentioned if there is color contrast. In Experiment 3, participants named a target color faster in monochrome than in polychrome displays, suggesting that the effect of color contrast is not analogous to ease of production. We conclude that the tendency to overspecify color in polychrome displays is not a bottom-up effect driven by the visual salience of color as a property, but possibly a learned communicative strategy. We discuss the implications of our account for pragmatic theories of referential communication and models of audience design, challenging the view that overspecification is a form of egocentric behavior.