Towards a Flexible Semantics: Colour Terms in Collaborative Reference Tasks

First Joint Conference on Lexical and Computational Semantics (*SEM), pages 80–84, Montr´eal, Canada, June 7-8, 2012. c⃝2012 Association for Computational Linguistics Towards a Flexible Semantics: Colour Terms in Collaborative Reference Tasks Bert Baumgaertner University of California, Davis bbaum@ucdavis.edu Raquel Fern´andez University of Amsterdam raquel.fernandez@uva.nl Matthew Stone Rutgers University matthew.stone@rutgers.edu Abstract We report ongoing work on the development of agents that can implicitly coordinate with their partners in referential tasks, taking as a case study colour terms. We describe algo- rithms for generation and resolution of colour descriptions and report results of experiments on how humans use colour terms for reference in production and comprehension. 1 Introduction Speakers do not always share identical semantic rep- resentations nor identical lexicons. For instance, a subject may refer to a shape as a diamond while another subject may call that same shape a square (which just happens to be tilted sidewise); or some- one may refer to a particular colour with ‘light pink’ while a different speaker may refer to it as ‘salmon’. Regardless of these differences, which seem com- mon place, speakers in dialogue are able to com- municate successfully most of the time. Success- ful communication exploits interlocutors’ abilities to negotiate referring expressions interactively through grounding (Clark and Wilkes-Gibbs, 1986; Clark and Schaefer, 1989), but in many cases interlocutors can already make a good guess at their partners’ in- tentions by relaxing the interpretation of their utter- ances and looking for the referent that best matches this looser interpretation. We are interested in mod- elling this second kind of behaviour computation- ally, to get a better understanding of it and to con- tribute to the development of dialogue systems that are able to better coordinate with their human part- ners. In this paper we focus on collaborative referen- tial tasks (akin to the classic matching tasks intro- duced by Krauss and Weinheimer (1966) and Clark and Wilkes-Gibbs (1986)) and take as a case study colour terms. Our focus here is not on the explicit joint negotiation of effective terms, but rather on the deployment of flexible semantic representations that can adapt to the constraints imposed by the context and to the dialogue partner’s language use. We start by describing our algorithms for genera- tion and resolution of colour descriptions in the next section. In sections 3 and 4, we present results of experiments that investigate how humans use colour terms for reference in production and comprehen- sion. Section 5 compares our model against the ex- perimental data we have collected so far and dis- cusses some directions for future work. We end with a short conclusion in section 6. 2 Reference to Colours: Our Model Our view of how colour terms are used in referential tasks follows the basic tenets of Gricean pragmat- ics (Grice, 1975) and collaborative reference theo- ries (Clark and Wilkes-Gibbs, 1986), according to which speakers and addressees tend to maximize the success of their joint task while minimizing costs. In the domain of colour terms, we take this …

Content truncated to 500 words