Verb Phrase Ellipsis Resolution Using Discriminative and Margin-Infused Algorithms

Abstract

Verb Phrase Ellipsis (VPE) is an anaphoric construction in which a verb phrase has been elided. It occurs frequently in dialogue and informal conversational settings, but despite its evident impact on event coreference resolution and extraction, there has been relatively little work on computational methods for identifying and resolving VPE. Here, we present a novel approach to detecting and resolving VPE by using supervised discriminative machine learning techniques trained on features extracted from an automatically parsed, publicly available dataset. Our approach yields state-of-the-art results for VPE detection by improving F1 score by over 11%; additionally, we explore an approach to antecedent identification that uses the Margin-Infused-RelaxedAlgorithm, which shows promising results.

Improving Open-Domain Dialogue Systems via Multi-Turn Incomplete Utterance Restoration

“…Kumar and Joshi (2016) approach NSU resolution as a Seq2Seq learning problem. One type of NSU that draws extra attention is ellipsis (Merchant, 2016;Kenyon-Dean et al, 2016;McShane and Babkin, 2016).…”

A Neural Network Approach to Verb Phrase Ellipsis Resolution

“…Cheung, and Precup 2016). Previous work on VPE resolution has addressed the sub-tasks of VPE detection (Hardt 1992;1997;Hobbs and Kehler 1997;Nielsen 2003b;2004a;2004b) and VPE resolution (Hardt 1998;Nielsen 2003a;Bos 2005) separately, pipeline processing of VPE detection and VPE resolution (Bos and Spenader 2011;Bos 2012;Kawai 2013;Liu, Pellicer, and Gillick 2016;Bakhshandeh, Wellwood, and Allen 2016), and end-to-end modeling of the two steps (Kenyon-Dean, Cheung, and Precup 2016;McShane and Babkin 2016). Most existing work uses heuristic rules, manually constructed features extracted from auxiliaries and syntactic structures, and linguistic theories such as Discourse Representation Structure (DRS) (Bos 2012), Simple Parallel Configuration (SPC) (McShane and Babkin 2016), etc 1 .…”

“…Previous research (Nielsen 2003b;2004a;2004b;Bos 2012;Liu, Pellicer, and Gillick 2016;Kenyon-Dean, Cheung, and Precup 2016) regards VPE detection as a binary classification task over auxiliary verbs. In this paper, we follow the task specification.…”

“…For VPE detection, the first two baselines include a rule-based approach (Rule) and a machine learning based approach (ML). Both are proposed by Kenyon-Dean, Cheung, and Precup (2016). The third baseline is a 3-step approach proposed by Liu, Pellicer, and Gillick (2016), which includes VPE detection, antecedent identification and VPE resolution.…”

“…For VPE resolution, the first baseline is a DR theory-based VPE resolution approach (DRVPE) proposed by Bos (2012). The second is a Margin-Infused-Relaxed-Algorithm based approach (MIRA) for VPE resolution, which is proposed by Kenyon-Dean, Cheung, and Precup (2016). The third is the 3-step VPE resolution approach of Liu, Pellicer, and Gillick (2016).…”