publications

2023

1. Perspectival biscuits

   Karl Mulligan, and Kyle Rawlins

   In Proceedings of Semantics and Linguistic Theory

   absract Preprint Handout

   This paper describes a novel class of biscuit conditional, the ‘perspectival biscuit’, which arises when an if-clause containing a generic pronoun (e.g., generic you) is used to shift perspective for the interpretation of a perspective-sensitive item in the consequent: e.g., fixing the directionality of behind in “If you’re at the door, the cat is behind the desk.” This sentence is like a biscuit conditional in that it entails a fully-specified, propositionally stable consequent describing the spatial configuration of cat and desk, but this reading vanishes in favor of a conditional dependence reading when the antecedent contains any non-generic DP, a prediction that is not straightforwardly accounted for by existing theories of biscuit conditionals. An analysis is given demonstrating that biscuithood for perspectival biscuits arises due to generic quantification exclusively over individuals, not worlds.

2022

1. A compositional semantics for spatial perspective-shifting adjuncts

   Karl Mulligan, and Kyle Rawlins

   In Proceedings of the West Coast Conference on Formal Linguistics

   absract Preprint

   This paper gives a compositional account of a construction which shifts perspective in projective prepositions such as behind and left of: the left-adjoined from-adjunct, as in "From the door, the ball is behind the chair." The implicit argument for perspective is normally assigned its value anaphorically, but in case of shifting with from-adjuncts, this variable is instead bound. The binding operator introduced by left-adjoined from is analyzed as a generic quantifier over simulated individuals, in the sense of Moltmann (2006), with the locational information in the from-adjunct forming the content of the restrictor. This paper also analyzes perspective shifting with a truth-conditionally equivalent expression, the conditional paraphrase "If you’re at the door," using a similar approach.

2021

1. Structure here, bias there: hierarchical generalization by jointly learning syntactic transformations

   Karl Mulligan, Robert Frank, and Tal Linzen

   In Proceedings of the Society for Computation in Linguistics

   absract Link PDF

   When learning syntactic transformations, children consistently induce structure-dependent generalizations, even though the primary linguistic data may be consistent with both linear and hierarchical rules. What is the source of this inductive bias? In this paper, we use computational models to investigate the hypothesis that evidence for the structure-sensitivity of one syntactic transformation can bias the acquisition of another transformation in favor of a hierarchical rule. We train sequence-to-sequence models based on artificial neural networks to learn multiple syntactic transformations at the same time in a fragment of English; we hold out cases that disambiguate linear and hierarchical rules for one of those transformations, and then test for hierarchical generalization to these held-out sentence types. Consistent with our hypothesis, we find that multitask learning induces a hierarchical bias for certain combinations of tasks, and that this bias is stronger for transformations that share computational building blocks. At the same time, the bias is in general insufficient to lead the learner to categorically acquire the hierarchical generalization for the target transformation.

2020

1. Grounded sequence to sequence transduction

   Lucia Specia, Loic Barrault, Ozan Caglayan, and 15 more authors

   absract Link PDF

   Speech recognition and machine translation have made major progress over the past decades, providing practical systems to map one language sequence to another. Although multiple modalities such as sound and video are becoming increasingly available, the state-of-the-art systems are inherently unimodal, in the sense that they take a single modality - either speech or text - as input. Evidence from human learning suggests that additional modalities can provide disambiguating signals crucial for many language tasks. In this article, we describe the How2 dataset, a large, open-domain collection of videos with transcriptions and their translations. We then show how this single dataset can be used to develop systems for a variety of language tasks and present a number of models meant as starting points. Across tasks, we find that building multimodal architectures that perform better than their unimodal counterpart remains a challenge. This leaves plenty of room for the exploration of more advanced solutions that fully exploit the multimodal nature of the How2 dataset, and the general direction of multimodal learning with other datasets as well.