Parsing the WSJ Using CCG and Log-Linear Models

By Stephen Clark, James Richard Curran. Annual Meeting of the Association for Computational Linguistics, 2004

352 citations recorded by the source corpus.

This paper describes and evaluates log-linear parsing models for Combinatory Categorial Grammar (CCG). A parallel implementation of the L-BFGS optimisation algorithm is described, which runs on a Beowulf cluster allowing the complete Penn Treebank to be used for estimation. We also develop a new efficient parsing algorithm for CCG which maximises expected recall of dependencies. We compare models which use all CCG derivations, including non-standard derivations, with normal-form models. The performances of the two models are comparable and the results are competitive with existing wide-coverage CCG parsers.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/b9eea85e590f6e522e3681b8e45012684c60b0fd

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

Programming With Unrestricted Natural Language (2005, methodology): “We use the C&C CCG parser (Clark and Curran, 2004) for this first stage of processing.”
Probabilistic CFG with Latent Annotations (2005, methodology): “Several parsing algorithms that also use inside-outside calculation on packed chart have been proposed (Goodman, 1996b; Sima´an, 2003; Clark and Curran, 2004).”
Bootstrapping language acquisition. (2017, methodology): “CCG has been extensively applied to parsing and interpretation of unrestricted text (Auli & Lopez, 2011; Clark & Curran, 2004; Hockenmaier, 2003; Lewis & Steedman, 2014), and has received considerable attention recently in the computational literature on semantic parser induction (e.g., Artzi, Das,…”
Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models (2007, methodology): “We also showed in Clark and Curran (2004b) that, in contrast with Hockenmaier (2003a), adding distance to the dependency features in the log-linear model does improve parsing accuracy.”
Large-scale Semantic Parsing without Question-Answer Pairs (2014, methodology): “For these reasons, we rely on the C&C parser (Clark and Curran, 2004), a general-purpose CCG parser, to obtain syntactic derivations.”
Question classification with log-linear models (2006, methodology): “Each question was parsed using the C&C CCG parser [1] with a model speciﬁcally created for parsing questions.”
Toponym resolution in text: annotation, evaluation and applications of spatial grounding (2007, neutral): “C&C (Curran and Clark (2003a,b); Clark and Curran (2004)) is an NLP toolkit for sequence tagging and parsing.”
Linguistically Motivated Large-Scale NLP with C&C and Boxer (2007, methodology): “The combination of ﬁnite-state supertagging and highly engineered C++ leads to a parser which can analyse up to 30 sentences per second on standard hardware (Clark and Curran, 2004a).”
Combining Symbolic and Distributional Models of Meaning (2007, background): “However, the recent improvements in wide-coverage parsing (Clark & Curran 2004) have led to attempts to scale up symbolic approaches, with some success (Bos 2005).”
A Natural Proof System for Natural Language (2017, methodology): “The C&C-based prover fails in the beginning when the parser fails to return the derivations for the first two sentences, which contain relevant information for the entailment.”
An Introduction to Conditional Random Fields (2010, methodology): “Also, there is work on using CRFs to learn distributions over the derivations of a grammar [99, 19, 127, 31].”
Learning syntactic structure (2007, neutral): “Even in a grammatical framework such as combinatory categorial grammar (Steedman 2000), which is supposed to easily accommodate incremental parsing, wide coverage parsers (Hockenmaier and Steedman 2002; Clark and Curran 2004) are not incremental.”
An Introduction to Conditional Random Fields for Relational Learning (2007, neutral): “Also, there is work on using CRFs to learn distributions over the derivations of a grammar [Riezler et al., 2002, Clark and Curran, 2004, Sutton, 2004, Viola and Narasimhan, 2005].”
Wide-Coverage Semantic Analysis with Boxer (2008, methodology): “The existence of CCGbank (Hockenmaier, 2003) and robust parsers trained on it (Clark and Curran, 2004; Bos et al., 2004) make Boxer a state-of-the-art open-domain tool for deep semantic analysis.”
Online Large-Margin Training of Dependency Parsers (2005, background): “Discriminatively trained parsers that score entire trees for a given sentence have only recently been investigated (Riezler et al., 2002; Clark and Curran, 2004; Collins and Roark, 2004; Taskar et al., 2004).”