Year: 2016
Type: conference-paper
Abstract: We consider the task of learning a contextdependent mapping from utterances to denotations.With only denotations at training time, we must search over a combinatorially large space of logical forms, which is even larger with context-dependent utterances.To cope with this challenge, we perform successive projections of the full model onto simpler models that operate over equivalence classes of logical forms.Though less expressive, we find that these simpler models are much faster and can be surprisingly effective.Moreover, they can be used to bootstrap the full model.Finally, we collected three new contextdependent semantic parsing datasets, and develop a new left-to-right parser.Property[p] Value[v] ⇒ Set[p(v)] all entities whose property p is v Set[s] Property[p] ⇒ Value[argmin/argmax(s, p)] element in s with smallest/largest p Set[s] Int[i] ⇒ Value[s[i]] i-th element of s Action[a] Value[v1] Value[v2] ⇒ Root[a(v1, v2)] top-level action applied to arguments v1, v2
DOI: 10.18653/v1/p16-1138
Authors: Reginald Long, Panupong Pasupat, Percy Liang
Institution: Stanford University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Funder: Microsoft Research
OpenAlex ID: W2963372003