A Neural Network Approach to Verb Phrase Ellipsis Resolution

Abstract

Verb Phrase Ellipsis (VPE) is a linguistic phenomenon, where some verb phrases as syntactic constituents are omitted and typically referred by an auxiliary verb. It is ubiquitous in both formal and informal text, such as news articles and dialogues. Previous work on VPE resolution mainly focused on manually constructing features extracted from auxiliary verbs, syntactic trees, etc. However, the optimization of feature representation, the effectiveness of continuous features and the automatic composition of features are not well addressed. In this paper, we explore the advantages of neural models on VPE resolution in both pipeline and end-to-end processes, comparing the differences between statistical and neural models. Two neural models, namely multi-layer perception and the Transformer, are employed for the subtasks of VPE detection and resolution. Experimental results show that the neural models outperform the state-of-the-art baselines in both subtasks and the end-to-end results.

Are Ellipses Important for Machine Translation?

“…We use manually reconstructed, gold sentences to analyse the exact impact of this procedure on MT. In practice, the text can be fed into an ellipses resolution system (Khullar 2020;Zhang et al 2019) as a preprocessing step. This means that the accuracy of such a system will also contribute to the final translation quality.…”

“…Ellipsis has been thoroughly studied in theoretical linguistics (Halliday and Hasan 1976;Hankamer 1978;Lobeck 1995;Merchant 2004Merchant , 2010Gunther 2011;van Craenenbroeck and Merchant 2013;Miller and Pullum 2013;Park 2017), in cognitive linguistics (Kim, Brehm, and Yoshida 2019), and in language acquisition studies (Hyams, Mateu, and Winans 2017;Lindenbergh, van Hout, and Hollebrandse 2015;Goksun et al 2010;Wijnen, Roeper, and van der Meulen 2003). Previous computational work on ellipsis resolution has mostly focused on Verb Phrase Ellipsis (VPE), gapping and sluicing; for instance, the detection of VPE in the Penn Treebank using pattern match (Hardt 1992), a transformation learning-based approach to generated patterns for VPE resolution (Hardt 1998), the domain independent VPE detection and resolution using machine learning (Nielsen 2003), automatically parsed text (Nielsen 2004), sentence trimming methods (McShane, Nirenburg, and Babkin 2015), linguistic principles (McShane and Babkin 2016), improved parsing techniques that encode elided material dependencies for reconstruction of sentences containing gapping (Schuster, Nivre, and Manning 2018), discriminative and margin infused algorithms (Dean, Cheung, and Precup 2016), Multilayer Perceptrons (MLP) and Transformers (Zhang et al 2019). Computational work on noun ellipsis is comparatively sparse, comprising a simple rule based system (Khullar, Anthony, and Shrivastava 2019), an annotated corpus for noun ellipsis in movie dialogues (Khullar, Majmundar, and Shrivastava 2020), and end-to-end resolution pipeline experiments with statistical and neural model experiments (Khullar 2020).…”

“…Ellipses are not very frequent in text, 4 but for improving the accuracy of Natural Language Processing (NLP) systems that handle data with ellipses, they are important (Zhang et al 2019;Dean, Cheung, and Precup 2016). One such NLP application could be Machine Translation (MT).…”