Paper: Structured Composition of Semantic Vectors

ACL ID W11-0131
Title Structured Composition of Semantic Vectors
Venue IWCS
Session Main Conference
Year 2011

Distributed models of semantics assume that word meanings can be discovered from “the com- pany they keep.” Many such approaches learn semantics from large corpora, with each document considered to be unstructured bags of words, ignoring syntax and compositionality within a docu- ment. In contrast, this paper proposes a structured vectorial semantic framework, in which semantic vectors are defined and composed in syntactic context. As such, syntax and semantics are fully interactive; composition of semantic vectors necessarily produces a hypothetical syntactic parse. Evaluations show that using relationally-clustered headwords as a semantic space in this framework improves on a syntax-only model in perplexity and parsing accuracy.