Accurate Unlexicalized Parsing

By Daniel Klein, Christopher D. Manning. 2003

3,472 citations recorded by the source corpus.

We demonstrate that an unlexicalized PCFG can parse much more accurately than previously shown, by making use of simple, linguistically motivated state splits, which break down false independence assumptions latent in a vanilla treebank grammar. Indeed, its performance of 86.36% (LP/LR F1) is better than that of early lexicalized PCFG models, and surprisingly close to the current state-of-the-art. This result has potential uses beyond establishing a strong lower bound on the maximum possible accuracy of unlexicalized models: an unlexicalized PCFG is much more compact, easier to replicate, and easier to interpret than more complex lexical models, and the parsing algorithms are simpler, more widely understood, of lower asymptotic complexity, and easier to optimize.

Subfields: Artificial Intelligence, Natural Language Processing

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

In Natural Language Processing

Citations in the corpus

Grammar as a Foreign Language (2014, background): “The focus in these approaches is on devising appropriate smoothing techniques for highly lexicalized and thus rare events (Collins, 1997) or ca efully crafting the model structure (Klein & Manning, 2003).”
A large annotated corpus for learning natural language inference (2015, methodology): “2 (Klein and Manning, 2003), trained on the standard training set as well as on the Brown Corpus (Francis and Kucera 1979), which we found to improve the parse quality of the descriptive sentences and noun phrases found in the descriptions.”
Distributed Representations of Sentences and Documents (2014, methodology): “To achieve this, Socher et al. (Socher et al., 2013b) used the Stanford Parser (Klein & Manning, 2003) to parse each sentence to subphrases.”
Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank (2013, methodology): “It was parsed with the Stanford parser (Klein and Manning, 2003) and includes a total of 215,154 unique phrases from those parse trees, each annotated by 3 human judges.”
Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference (2019, methodology): “These parse templates are based on the parses generated by the Stanford PCFG Parser 3.5.2 (Klein and Manning, 2003 words, subsequences, and constituents).”
What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties (2018, methodology): “We parse them with the Stanford Parser (2017-06-09 version), using the pre-trained PCFG model (Klein and Manning, 2003), and we rely on the part-of-speech, constituency and dependency parsing information provided by this tool where needed.”
Neural Module Networks (2015, methodology): “Parsing We begin by parsing each question with the Stanford Parser [19].”
SPICE: Semantic Propositional Image Caption Evaluation (2016, methodology): “A Probabilistic Context-Free Grammar (PCFG) dependency parser [19] is followed by three post-processing steps that simplify quantiﬁcational modiﬁers, resolve pronouns and handle plural nouns.”
Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks (2015, neutral): “3.2) on binarized constituency trees (Klein and Manning, 2003).”
Enhanced LSTM for Natural Language Inference (2016, background): “3 (Klein and Manning, 2003) and they are delivered as part of the SNLI corpus.”
Exploring Models and Data for Image Question Answering (2015, methodology): “We used the Stanford parser (Klein & Manning, 2003) to obtain the syntatic structure of the original image description.”
Inferring and Executing Programs for Visual Reasoning (2017, neutral): “The main difference between our models and existing module networks is that we replace hand-designed off-the-shelf syntactic parsers [23], which perform poorly on complex questions such as those in CLEVR [19], by a learnt program generator that can adapt to the task at hand.”
Generating Typed Dependency Parses from Phrase Structure Parses (2006, methodology): “The Stanford system exploits the information about predicate-argument structure encoded in the generated typed dependency trees in three ways: in generating a quasi-logical representation of the event structure represented by each sentence (following the work of Moldovan and Harabagiu in question an”
TreeTalk: Composition and Compression of Trees for Image Descriptions (2014, methodology): “, 2011) parsed using the St nfo d p rser (Klein and Manning, 2003).”
Semantic Parsing via Paraphrasing (2014, methodology): “We run all questions through the Stanford CoreNLP pipeline (Toutanova and Manning, 2003; Finkel et al., 2005; Klein and Manning, 2003).”
What Does BERT Learn about the Structure of Language? (2019, background): “Each word’s role is computed based on its left-to-right index, its right-to-left index, an ordered pair containing its left-to-right and right-to-left indices, its position in a syntactic tree (formatted version of the Stanford PCFG Parser (Klein and Manning, 2003) with no unary nodes and no labels)”
Jointly Modeling Deep Video and Compositional Text to Bridge Vision and Language in a Unified Framework (2015, methodology): “First, we use the Stanford Parser (Klein and Manning 2013) to parse all sentences, then we choose nsubj to find subject-verb pair, and choose dobj to find verb-object pairs in order to obtain SV O triplets.”