Hybrid Grammars for Parsing of Discontinuous Phrase Structures and Non-Projective Dependency Structures

Abstract

We explore the concept of hybrid grammars, which formalize and generalize a range of existing frameworks for dealing with discontinuous syntactic structures. Covered are both discontinuous phrase structures and non-projective dependency structures. Technically, hybrid grammars are related to synchronous grammars, where one grammar component generates linear structures and another generates hierarchical structures. By coupling lexical elements of both components together, discontinuous structures result. Several types of hybrid grammars are characterized. We also discuss grammar induction from treebanks. The main advantage over existing frameworks is the ability of hybrid grammars to separate discontinuity of the desired structures from time complexity of parsing. This permits exploration of a large variety of parsing algorithms for discontinuous structures, with different properties. This is confirmed by the reported experimental results, which show a wide variety of running time, accuracy, and frequency of parse failures.

Advances in Using Grammars with Latent Annotations for Discontinuous Parsing

“…Vijay-Shanker et al (1987) and Gebhardt et al (2017) give formal definitions of these formalisms, respectively. Baseline grammars are induced from the training data using the induction techniques of Kallmeyer and Maier (2013) and Gebhardt et al (2017), respectively. LCFRS are either binarized right-to-left (r2 ) or head-outward (ho) with vertical and horizontal Markovization set to 1.…”

“…We consider models based on LCFRS and hybrid grammars. Vijay-Shanker et al (1987) and Gebhardt et al (2017) give formal definitions of these formalisms, respectively. Baseline grammars are induced from the training data using the induction techniques of Kallmeyer and Maier (2013) and Gebhardt et al (2017), respectively.…”

Constituency Parsing as an Instance of the M-monoid Parsing Problem

“…Instead, we introduce partitioned constituent trees which are inspired by Nederhof and Vogler (2014) [14] (also cf. [5]). They are tuples consisting of a (usual) tree, a strict total order on its leaves, and a partitioning of its leaves.…”

Discontinuous Grammar as a Foreign Language

“…These structures decompose the sentence into constituents (also called phrases) and establish hierarchical relations between them and the sentence's words, resulting in a tree structure. While regular (or continuous) constituent trees (as the one depicted in Figure 1(a)) are enough for representing a wide range of syntactic structures, it is necessary to use discontinuous constituent trees to fully describe all linguistic phenomena present in human languages (Gebhardt, Nederhof and Vogler, 2017). Although producing the latter is considered an especially challenging problem in constituent parsing (Corro, 2020), they are necessary for adequately representing some syntactic phenomena that occur in almost the 20% of the sentences from the most widely-used syntacticallyannotated corpus of English, the Penn Treebank (Marcus, Santorini and Marcinkiewicz, 1993) such as cross-serial dependencies, dislocations, long-distance extractions and some wh-movements (Evang and Kallmeyer, 2011), which require constituents with discontinuous spans and result in phrase-structure trees with crossing branches (as can be seen in Figure 1(b)).…”