Quasi-Synchronous Grammars: Alignment by Soft Projection of Syntactic Dependencies

By David A. Smith, Jason Eisner. 2006

113 citations recorded by the source corpus.

Many syntactic models in machine translation are channels that transform one tree into another, or synchronous grammars that generate trees in parallel. We present a new model of the translation process: quasi-synchronous grammar (QG). Given a source-language parse tree T1, a QG defines a monolingual grammar that generates translations of T1. The trees T2 allowed by this monolingual grammar are inspired by pieces of substructure in T1 and aligned to T1 at those points. We describe experiments learning quasi-synchronous context-free grammars from bitext. As with other monolingual language models, we evaluate the cross-entropy of QGs on unseen text and show that a better fit to bilingual data is achieved by allowing greater syntactic divergence. When evaluated on a word alignment task, QG matches standard baselines.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Graph Convolutional Encoders for Syntax-aware Neural Machine Translation (2017, neutral): “This goal is in line with claims that rigid syntactic constraints typically hurt MT (Zollmann and Venugopal, 2006; Smith and Eisner, 2006; Chiang, 2010), and, though these claims have been made in the context of traditional MT systems, we believe they are no less valid for NMT.”
Deep Learning for Answer Sentence Selection (2014, background): “[20] built a generative model to match the dependency trees of question answer pairs based on the soft alignment of a quasisynchronous grammar [16].”
Sentence Simplification by Monolingual Machine Translation (2012, methodology): “Woodsend and Lapata’s (2011) model is based on quasi-synchronous grammar (Smith and Eisner, 2006).”
Phrase Dependency Machine Translation with Quasi-Synchronous Tree-to-Tree Features (2014, methodology): “Smith and Eisner (2006) noted that these limitations of synchronous grammars result from an emphasis on generating the two strings.”
Sequence-to-Sequence Learning with Latent Neural Grammars (2021, unknown): “First, we model the distribution over the target sequence with a quasi-synchronous grammar [103] which assumes a hierarchical generative process whereby each node in the target tree is transduced by nodes in the source tree.”
Answer Extraction as Sequence Tagging with Tree Edit Distance (2013, methodology): “Prior work in this includes that of: Punyakanok et al. (2004), based on mapping syntactic dependency trees; Wang et al. (2007) utilizing Quasi-Synchronous Grammar (Smith and Eis-ner, 2006); Heilman and Smith (2010) using TED; and Shima et al. (2008), Ding et al. (2008) and Wang and Manning (2010), w”
Compositional Generalization and Natural Language Variation: Can a Semantic Parsing Approach Handle Both? (2020, methodology): “Notably, QCFGs cannot directly represent relations between source strings, such as semantic similarity, or relations between target strings, such as logical equivalence (e.g. intersect(a,b) ⇔ intersect(b,a) ), that could enable greater generalization.”
Paraphrase Identification as Probabilistic Quasi-Synchronous Recognition (2009, methodology): “Since it loosely links the two sentences’ syntactic structures, QG is well suited for problems like word alignment for MT (Smith and Eisner, 2006) and question answering (Wang et al., 2007).”
Adversarial Example Generation with Syntactically Controlled Paraphrase Networks (2018, methodology): “Most related to our methods are those based on syntactic transformations, like the tree-to-tree sentence simplification method of Woodsend and Lapata (2011) based on quasi-synchronous grammar (Smith and Eisner, 2006).”
Tree Edit Models for Recognizing Textual Entailments, Paraphrases, and Answers to Questions (2010, background): “The other system, by Das and Smith (2009), is based on a quasi-synchronous grammar (QG; Smith and Eisner, 2006), a probabilistic model that allows loose alignments between trees but prefers tree isomorphism.”
Title Generation with Quasi-Synchronous Grammar (2010, methodology): “In our experiments, we obtain this combined representation from the output of the Stanford parser (Klein and Manning, 2003) but any other broadly similar parser could be used instead.”
Question Answering Using Enhanced Lexical Semantic Models (2013, methodology): “The mapping is learned by a generative probabilistic model based on a Quasi-synchronous Grammar formulation (Smith and Eisner, 2006).”