Coarse-to-Fine n-Best Parsing and MaxEnt Discriminative Reranking

By Eugene Charniak, Mark Johnson. Annual Meeting of the Association for Computational Linguistics, 2005

1,293 citations recorded by the source corpus.

Discriminative reranking is one method for constructing high-performance statistical parsers (Collins, 2000). A discriminative reranker requires a source of candidate parses for each sentence. This paper describes a simple yet novel method for constructing sets of 50-best parses based on a coarse-to-fine generative parser (Charniak, 2000). This method generates 50-best lists that are of substantially higher quality than previously obtainable. We used these parses as the input to a MaxEnt reranker (Johnson et al., 1999; Riezler et al., 2002) that selects the best parse from the set of parses for each sentence, obtaining an f-score of 91.0% on sentences of length 100 or less.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/0ecb33ced5b0976accdf13817151f80568b6fdcb

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

Better k-best Parsing (2005, methodology): “…2003) with a beam width of 10−3 together with a cell limit of 100 to obtain k-best lists; the average number of parses obtained per sentence was 29.2, the maximum, 101.7 Charniak and Johnson (2005) use coarse-to-fine parsing on top of the Charniak (2000) parser and get 50-best lists for section 23.”
Unbounded Dependency Recovery for Parser Evaluation (2009, methodology): “McClosky et al. (2006) report an F-score of 92.1% using self-training applied to the reranker of Charniak and Johnson (2005).”
Reranking and Self-Training for Parser Adaptation (2006, methodology): “In particular, we show that the reranking parser described in Charniak and Johnson (2005) improves performance of the parser on Brown to 85.2%.”
Comparing and Combining Finite-State and Context-Free Parsers (2005, methodology): “Also, features of the sort used in our shallow parser could be included in a reranker, such as that in Charniak and Johnson (2005), with a context-free parsing accuracy objective.”
Ambiguity-aware Ensemble Training for Semi-supervised Dependency Parsing (2014, background): “…when employing more diverse parsers to compose the parse forest of higher quality for the unlabeled data, such as the easy-ﬁrst non-directional dependency parser (Goldberg and Elhadad, 2010) and other constituent parsers (Collins and Koo, 2005; Charniak and Johnson, 2005; Finkel et al., 2008).”
Bayesian Symbol-Refined Tree Substitution Grammars for Syntactic Parsing (2012, background): “…2009; Bansal and Klein, 2010) has consistently shown that a probabilistic TSG (PTSG) parser is signiﬁcantly more accurate than a PCFG parser, but is still inferior to state-of-the-art parsers (e.g., the Berkeley parser (Petrov et al., 2006) and the Charniak parser (Charniak and Johnson, 2005)).”
Inducing Tree-Substitution Grammars (2010, neutral): “0 Reranking parser (Charniak and Johnson, 2005) 92.”
From Baby Steps to Leapfrog: How “Less is More” in Unsupervised Dependency Parsing (2010, background): “…approach (see§5), it would be prudent to re-test Baby Steps with a lexicalized model. approaches to parsing, translation and speech recognition (Charniak and Johnson, 2005; Charniak et al., 2006; Petrov et al., 2008; Petrov, 2009), and recently unsupervised POS tagging (Ravi and Knight,…”
A Latent Variable Model for Generative Dependency Parsing (2007, background): “All the most accurate dependency parsing models are fully discriminative, unlike constituent parsing where all the state of the art methods have a genera- tive component (Charniak and Johnson, 2005; Henderson, 2004; Collins, 2000).”
CoNLL-2011 Shared Task: Modeling Unrestricted Coreference in OntoNotes (2011, methodology): “Parse Trees Predicted parse trees were produced using the Charniak parser (Charniak and Johnson, 2005).12 Some additional tag types used in the OntoNotes trees were added to the parser’s tagset, including the NML tag that has recently been added to capture internal NP structure, and the rules used…”
Measuring Syntactic Complexity in L2 Writing Using Fine‐Grained Clausal and Phrasal Indices (2018, background): “In recent years,1 however, technological advancements in the field of natural language processing (e.g., Charniak & Johnson, 2005; Chen & Manning, 2014; Klein & Manning, 2003) have paved the way for the accurate automatic annotation of linguistic features.”
Products of Random Latent Variable Grammars (2010, result): “These parsing results are even better than those obtained by discriminative systems which have access to additional non-local features (Charniak and Johnson, 2005; Huang, 2008).”
Dependency Parsing (2009, neutral): “For my experiments I will make use of Charniak and Johnson [2005] as well as other constituent parsers.”
A discriminative language model with pseudo-negative samples (2007, methodology): “We examined sentences using a phrase structure parser (Charniak and Johnson, 2005) and an HPSG parser 1 Since the PLM also made use of the BNC-corpus for positive examples, we were not able to classify sentences based on word occurrences (Miyao and Tsujii, 2005).”
Judging Grammaticality: Experiments in Sentence Classification (2013, methodology): “The Charniak and Johnson (2005) parser is used to produce the parse trees.”
From POS tagging to dependency parsing for biomedical event extraction (2018, background): “first row shows scores of BLLIP+Bio, the BLLIP reranking constituent parser (Charniak and Johnson, 2005) with an improved self-trained biomedical parsing model (McClosky, 2010).”