Abstract
We present a novel approach for generating effective referring expressions (REs). We define a synchronous grammar formalism that relates surface strings with the sets of objects they describe through an abstract syntactic structure. The grammars may choose to require or not that REs are distinguishing. We then show how to compute a chart that represents, in finite space, the complete (possibly infinite) set of valid REs for a target object. Finally, we propose a probability model that predicts how the listener will understand the RE, and show how to compute the most effective RE according to this model from the chart.
Highlights
The fundamental challenge in the generation of referring expressions (REG) is to compute an RE which is effective, i.e. understood as intended by the listener
These regions are distinguished by the semantic indices in the nonterminals that derive them; e.g., in Fig. 6, the subtree for “the square button” is an attempt to refer to b2, whereas the RE as a whole is meant to refer to b1
Based on an algorithm for computing a chart of all valid REs, we showed how to compute the RE that maximizes the probability of being understood as the target referent
Summary
The fundamental challenge in the generation of referring expressions (REG) is to compute an RE which is effective, i.e. understood as intended by the listener. We present an algorithm that generates the RE t that maximizes P (a⇤|t), i.e. the RE that has the highest chance to be understood correctly by the listener according to the probabilistic RE resolution model. This is a challenging problem, since the algorithm must identify that. We start by defining a synchronous grammar formalism that relates surface strings to their interpretations as sets of objects in a given domain (Section 3) This formalism integrates REG with surface realization, and allows us to specify in the grammar whether REs are required to be distinguishing. Proceedings of the INLG and SIGDIAL 2014 Joint Session, pages 6–15, Philadelphia, Pennsylvania, 19 June 2014. c 2014 Association for Computational Linguistics with probabilistic listener models
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.