---
title: "Syntax & grammar formalisms — papers — Elamur Research"
description: "Syntax & grammar formalisms: 162 papers of the Elamur Research library, the most cited first."
canonical: "https://research.elamur.ai/free/en/topics/syntax-grammar/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/topics/syntax-grammar/)

[home](https://md.research.elamur.ai/free/en/) / [topics](https://md.research.elamur.ai/free/en/topics/) / Syntax & grammar formalisms

# 📐 Syntax & grammar formalisms

162 papers · page 1 of 3 · 1–25 · the most cited first · [all topics](https://md.research.elamur.ai/free/en/topics/)

Constituency, categorial and feature-based grammars, and the parsers built on them.

💡 Explained: [Construction grammar](https://md.research.elamur.ai/free/en/concepts/construction-grammar/) · [Context-free grammar](https://md.research.elamur.ai/free/en/concepts/context-free-grammar/)

A survey · about 18 minutes to read · every paper named here has its own page in this library

**In this article**

1. [What it is](#a-what-it-is)
2. [History and milestones](#a-history-and-milestones)
3. [Main problems and tasks](#a-main-problems-and-tasks)
4. [Methods and schools](#a-methods-and-schools)
5. [Datasets, resources and benchmarks](#a-datasets-resources-and-benchmarks)
6. [Landmark papers in this library](#a-landmark-papers-in-this-library)
7. [People](#a-people)
8. [Institutions and places](#a-institutions-and-places)
9. [Open questions](#a-open-questions)
10. [Neighbouring topics](#a-neighbouring-topics)
11. [Key works in this library](#a-key-works-in-this-library)

Syntax is the part of grammar that governs how words combine into phrases and sentences, and a grammar formalism is a precise, usually mathematical, language for stating such rules. This topic gathers work on the major families of formalisms — phrase-structure (constituency) grammars, categorial grammars such as Combinatory Categorial Grammar (CCG), tree-adjoining grammars (TAG), and feature-based, constraint-based theories such as Lexical-Functional Grammar (LFG) and Head-Driven Phrase Structure Grammar (HPSG) — together with the parsers, treebanks and translation systems built on top of them.

The papers here come from two traditions that meet only occasionally. One is descriptive and theoretical linguistics: historical grammars of English written more than a century ago, minimalist analyses of Arabic dialects, and close studies of word order, negation and clause structure in Biblical Hebrew. The other is computational: how to assign structure to millions of sentences quickly and accurately, how to learn grammars from annotated corpora, how syntax can help machine translation, and, most recently, whether neural networks trained on raw text discover hierarchical structure on their own.

What unites them is a single question asked at different scales: what structure does a sentence have, and how can that structure be stated precisely enough to be checked, learned or computed? The sections below trace how the answers changed, from hand-written rule systems to treebank statistics and then to neural parsers and probing studies.

## What it is

A **constituency** (phrase-structure) analysis groups words into nested phrases — noun phrases, verb phrases, clauses — and labels each group. A **dependency** analysis instead links each word directly to the word it modifies. Most formalisms on this page are constituency-based or lexicalized: they place much of the grammar in rich lexical entries and use a small number of general combining operations. Dependency syntax as an annotation standard has its own pages, [Dependency parsing](https://md.research.elamur.ai/free/en/topics/dependency-parsing/) and [Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/), but the two views constantly feed each other, and conversion between them is a recurring theme here.

The main formalisms differ in what they take as primitive:

- **Context-free phrase-structure grammar (CFG)** rewrites one category into a sequence of categories. It is the backbone of treebanks such as the Penn Treebank and of most statistical constituency parsers.
- **Categorial grammars**, and CCG in particular, give every word a category that states what arguments it needs and on which side; a handful of combinatory rules (application, composition, type-raising) then assemble the sentence. Syntax and semantics are built in lockstep.
- **Tree-adjoining grammar** combines small elementary trees by substitution and by adjunction, an operation that inserts one tree into the middle of another. This lets a single lexical entry encode a word's whole argument frame.
- **LFG** separates a constituent structure (a tree) from a functional structure (an attribute–value matrix recording subject, object, tense and so on).
- **HPSG** describes every linguistic sign as a typed feature structure and states the grammar as constraints on such structures, with unification as the central operation.
- **Generative (transformational and minimalist) syntax**, in the tradition founded by Chomsky, is represented here mainly by descriptive studies of particular languages rather than by computational implementations.

A recurring theme is **formal power**. CFGs cannot describe some cross-serial dependencies found in natural languages, while unrestricted grammars are too powerful to parse efficiently. Much work therefore seeks a middle class of "mildly context-sensitive" formalisms that are expressive enough for natural language yet still parsable in polynomial time.

## History and milestones

Landmark papers on this topic: the 12 most cited, in time The 12 most cited papers on this topic span 2001–2019; the dot grows with the citations.

**The numbers**

| Paper | Citations |
| --- | --- |
| [2001 — A Syntax-based Statistical Translation Model](https://md.research.elamur.ai/free/en/papers/yamada-2001-syntax-based-statistical-translation-model/) | 1,019 |
| [2006 — Generating Typed Dependency Parses from Phrase Structure Parses](https://md.research.elamur.ai/free/en/papers/marneffe-2006-generating-typed-dependency-parses-phrase/) | 2,886 |
| [2007 — Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models](https://md.research.elamur.ai/free/en/papers/clark-2007-wide-coverage-efficient-statistical-parsing/) | 626 |
| [2007 — CCGbank: A Corpus of CCG Derivations and Dependency Structures Extracted from the Penn Treebank](https://md.research.elamur.ai/free/en/papers/hockenmaier-2007-ccgbank-corpus-ccg-derivations-dependency/) | 612 |
| [2008 — The Importance of Syntactic Parsing and Inference in Semantic Role Labeling](https://md.research.elamur.ai/free/en/papers/punyakanok-2008-importance-syntactic-parsing-inference-semantic/) | 538 |
| [2012 — Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars](https://md.research.elamur.ai/free/en/papers/zettlemoyer-2012-learning-map-sentences-logical-form/) | 1,074 |
| [2015 — Two/Too Simple Adaptations of Word2Vec for Syntax Problems](https://md.research.elamur.ai/free/en/papers/ling-2015-two-too-simple-adaptations-word2vec/) | 504 |
| [2016 — Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://md.research.elamur.ai/free/en/papers/linzen-2016-assessing-ability-lstms-learn-syntax/) | 1,243 |
| [2016 — Does String-Based Neural MT Learn Source Syntax?](https://md.research.elamur.ai/free/en/papers/shi-2016-does-string-based-neural-mt/) | 496 |
| [2018 — Constituency Parsing with a Self-Attentive Encoder](https://md.research.elamur.ai/free/en/papers/kitaev-2018-constituency-parsing-self-attentive-encoder/) | 762 |
| [2019 — A Structural Probe for Finding Syntax in Word Representations](https://md.research.elamur.ai/free/en/papers/hewitt-2019-structural-probe-finding-syntax-word/) | 1,232 |
| [2019 — Linguistic Knowledge and Transferability of Contextual Representations](https://md.research.elamur.ai/free/en/papers/liu-2019-linguistic-knowledge-transferability-contextual-representations/) | 1,015 |

### Descriptive grammar

The oldest works here are grammars of English from the philological tradition: [Streifzüge durch die mittelenglische Syntax](https://md.research.elamur.ai/free/en/papers/einenkel-1887-streifzuge-durch-die-mittelenglische/) by Eugen Einenkel, [Historical Outlines of English Syntax](https://md.research.elamur.ai/free/en/papers/kellner-1892-historical-outlines-english-syntax/) by Leon Kellner, [System der neuenglischen Syntax](https://md.research.elamur.ai/free/en/papers/deutschbein-1917-system-der-neuenglischen-syntax/) by Max Deutschbein and the syntax volume of George Curme's [A Grammar of the English Language, Vol. III: Syntax](https://md.research.elamur.ai/free/en/papers/curme-1931-grammar-english-language-vol/). They catalogue constructions and their history in prose, without a formal notation, and remain the kind of source later formal and corpus work tests itself against.

### Formal grammar and the mildly context-sensitive programme

In the late 1950s Noam Chomsky's generative grammar made syntax a matter of explicit rule systems and classified grammars by generative power. During the 1970s and 1980s several alternatives appeared: Aravind Joshi's tree-adjoining grammar, Kaplan and Bresnan's LFG, generalized phrase structure grammar, Pollard and Sag's HPSG, and Steedman's combinatory version of categorial grammar. A key paper of this era in the library is [Characterizing Structural Descriptions Produced by Various Grammatical Formalisms](https://md.research.elamur.ai/free/en/papers/shanker-1987-characterizing-structural-descriptions-produced-various/) (1987), which compared formalisms by the trees they build rather than by the strings they accept and introduced Linear Context-Free Rewriting Systems. Three years later, [Synchronous Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/shieber-1990-synchronous-tree-adjoining-grammars/) showed how pairing two grammars could carry syntax over to logical form or to another language.

### Treebanks and statistical parsing

The release of the Penn Treebank in the early 1990s changed the field. Parsers could now be trained and scored on the same tens of thousands of hand-bracketed sentences, and probabilistic CFG parsers by Collins, Charniak and others dominated the late 1990s and 2000s. Researchers working on richer formalisms responded by deriving their own resources from it or by attaching statistical models to hand-built grammars: CCGbank for CCG, treebank-trained disambiguation for HPSG and LFG, and the dynamic Redwoods treebank for an English HPSG grammar. The years 2005–2009 gather the largest share of citations in this library, reflecting this period of "deep" statistical parsing and of syntax-based statistical machine translation.

### Neural parsing and the probing era

From about 2015, recurrent and then self-attentive neural encoders replaced hand-crafted features. Span-based neural parsers made constituency parsing simpler and more accurate, and pretrained contextual representations raised scores again. At the same time a new research question appeared: if language models learn syntax implicitly, can it be found inside them? Studies of agreement in LSTMs and geometric probes of BERT made syntax a test of what neural models understand. Recent years have also seen renewed theoretical work — constraint-based surveys of Germanic syntax, meaning-representation frameworks and minimalist studies of Arabic dialects.

Papers on this topic by year 162 dated papers from 1887 to 2026 (4 before 1966 are not drawn); the busiest year is 2017 with 13, and 41% of them appeared in 2018 or later.

**The numbers**

| Year | Papers |
| --- | --- |
| 2026 | 12 |
| 2025 | 8 |
| 2024 | 1 |
| 2023 | 5 |
| 2022 | 3 |
| 2021 | 4 |
| 2020 | 10 |
| 2019 | 13 |
| 2018 | 10 |
| 2017 | 13 |
| 2016 | 9 |
| 2015 | 3 |
| 2014 | 2 |
| 2013 | 4 |
| 2012 | 5 |
| 2011 | 1 |
| 2010 | 7 |
| 2009 | 4 |
| 2008 | 4 |
| 2007 | 6 |
| 2006 | 5 |
| 2005 | 1 |
| 2004 | 5 |
| 2003 | 2 |
| 2002 | 4 |
| 2001 | 2 |
| 2000 | 1 |
| 1999 | 4 |
| 1998 | 3 |
| 1994 | 3 |
| 1991 | 1 |
| 1990 | 1 |
| 1988 | 1 |
| 1987 | 1 |
| 1931 | 1 |
| 1917 | 1 |
| 1892 | 1 |
| 1887 | 1 |

## Main problems and tasks

- **Constituency parsing**: assigning a labelled phrase-structure tree to a sentence, usually scored by bracket F1 on the newswire sections of the Penn Treebank.
- **Deep or formalism-specific parsing**: producing CCG derivations, HPSG analyses or LFG f-structures, which encode long-range dependencies, control and coordination that plain trees leave implicit.
- **Supertagging**: assigning each word a rich lexical category (a CCG category or a TAG elementary tree) before parsing; because lexicalized grammars push most decisions into the lexicon, this step does much of the work.
- **Disambiguation**: choosing among the many analyses a broad grammar allows, typically with log-linear (maximum entropy) models over packed parse forests.
- **Grammar engineering**: writing and maintaining large hand-built grammars, keeping them consistent across languages, and making them robust on real text.
- **Conversion between representations**: turning phrase-structure trees into typed dependencies, CCG derivations or other formalisms, and measuring what is lost.
- **Syntax in applications**: using trees in machine translation, semantic role labelling, paraphrase generation, error correction and coreference.
- **Parsing morphologically rich languages**, where words carry much of the grammatical information and segmentation must be decided together with syntax.
- **Evaluation across formalisms**, since parsers that output different structures cannot be compared on the same raw score.
- **Diagnosing neural models**: testing whether networks represent hierarchical structure, agreement and constituency.

## Methods and schools

### Hand-written grammars

The grammar-engineering school writes broad grammars by hand in a precise formalism and adds robustness and statistics afterwards. The English XTAG grammar, the multilingual LFG grammars of the ParGram project, the English Resource Grammar in HPSG and the Grammatical Framework (GF) resource grammars belong here. Their strength is precise, linguistically detailed output — including semantic representations — and their weakness is coverage on unrestricted text, which they address with fallback mechanisms such as fragment parsing and chunking.

### Treebank grammars and statistical disambiguation

A second approach extracts a grammar from an annotated corpus. CCGbank made it possible to train wide-coverage CCG parsers directly from Penn Treebank data, and similar conversions produced TAG and HPSG lexicons. Statistical models then choose among competing analyses. Feature forests and packed charts allowed log-linear models to be trained over exponentially many analyses without listing them, a technique shared by the CCG and HPSG parsers described below.

### Neural parsing

Neural parsers dropped explicit grammars in favour of scoring spans or transitions directly with learned encoders. A span-based chart parser with a BiLSTM encoder, and later with self-attention, reached state-of-the-art accuracy with very little linguistic machinery. In CCG, the most influential neural move was to put the whole statistical model into the supertagger and to search exactly with A*.

### Syntax for translation

Statistical machine translation went through a long phase of syntax-based models: tree-to-string channels, synchronous context-free grammars, hierarchical phrase-based models with syntactic labels, and tree-to-tree systems. In neural translation the question shifted to whether explicit source syntax still helps a model that can learn structure implicitly, and the answers in this library are mixed.

### Main debates

Several debates run through the topic. One is **depth versus robustness**: whether a linguistically precise grammar is worth its coverage cost. Another is **dependency versus constituency**, which shared tasks have found to matter less than often assumed. A third is **explicit versus implicit structure**: whether neural networks need trees built into their architecture or can learn hierarchical behaviour from strings alone. In theoretical linguistics, minimalist and constraint-based approaches continue to offer different accounts of the same constructions.

## Datasets, resources and benchmarks

- **Penn Treebank** (Wall Street Journal section): the standard English constituency benchmark since the 1990s; most parsing results here report F1 on its section 23. [Adding Noun Phrase Structure to the Penn Treebank](https://md.research.elamur.ai/free/en/papers/vadas-2007-adding-noun-phrase-structure-penn/) supplied the internal noun-phrase brackets it originally lacked.
- **CCGbank**: CCG derivations and word–word dependencies converted from the Penn Treebank, described in [CCGbank: A Corpus of CCG Derivations and Dependency Structures Extracted from the Penn Treebank](https://md.research.elamur.ai/free/en/papers/hockenmaier-2007-ccgbank-corpus-ccg-derivations-dependency/).
- **LinGO Redwoods**: a treebank of HPSG analyses built by selecting among parses of a hand-written grammar, see [LinGO Redwoods: A Rich and Dynamic Treebank for HPSG](https://md.research.elamur.ai/free/en/papers/oepen-2004-lingo-redwoods-rich-dynamic-treebank/).
- **SPMRL data**: constituency and dependency treebanks for nine morphologically rich languages, introduced through the shared task in [Overview of the SPMRL 2013 Shared Task](https://md.research.elamur.ai/free/en/papers/seddah-2013-overview-spmrl-2013-shared-task/).
- **Lexicalized grammars as resources**: the English XTAG grammar ([A Lexicalized Tree Adjoining Grammar for English](https://md.research.elamur.ai/free/en/papers/group-1998-lexicalized-tree-adjoining-grammar-english/)), [German and English Treebanks and Lexica for Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/kaeshammer-2012-german-english-treebanks-lexica-tree/) and the GF resource library.
- **DepBank and the British National Corpus** as independent evaluation material, used in [Formalism-Independent Parser Evaluation with CCG and DepBank](https://md.research.elamur.ai/free/en/papers/clark-2007-formalism-independent-parser-evaluation-ccg/) and [Parser Evaluation and the BNC](https://md.research.elamur.ai/free/en/papers/foster-2008-parser-evaluation-bnc-evaluating-4/).
- **Agreement test sets** built from Wikipedia for probing neural models, introduced in the LSTM agreement study discussed below.

## Landmark papers in this library

Citations of the papers on this topic, by when they were published Papers published in 2015–2019 gather 38% of the 24,866 citations of the papers on this topic.

**The numbers**

| Published | Papers and their citations |
| --- | --- |
| 2025–2029 | 20 papers, 16 citations |
| 2020–2024 | 23 papers, 508 citations |
| 2015–2019 | 48 papers, 9,326 citations |
| 2010–2014 | 19 papers, 2,513 citations |
| 2005–2009 | 20 papers, 6,781 citations |
| 2000–2004 | 14 papers, 3,753 citations |
| 1995–1999 | 7 papers, 757 citations |
| 1990–1994 | 5 papers, 777 citations |
| 1985–1989 | 2 papers, 435 citations |
| 1930–1934 | 1 papers, 0 citations |
| 1915–1919 | 1 papers, 0 citations |
| 1890–1894 | 1 papers, 0 citations |
| 1885–1889 | 1 papers, 0 citations |

### Typed dependencies from phrase structure (2006)

[Generating Typed Dependency Parses from Phrase Structure Parses](https://md.research.elamur.ai/free/en/papers/marneffe-2006-generating-typed-dependency-parses-phrase/) is the most cited work here. It defined what became known as Stanford Dependencies: a set of grammatical relations arranged in a hierarchy, obtained from Penn-style trees by head rules that favour content words, with an optional "collapsed" form that folds prepositions and conjunctions into the relation names. The scheme became a common output format for English parsers and a direct ancestor of Universal Dependencies, which is why a constituency paper sits at the root of so much dependency work.

### Can LSTMs learn agreement? (2016)

[Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://md.research.elamur.ai/free/en/papers/linzen-2016-assessing-ability-lstms-learn-syntax/) used English subject–verb agreement, especially sentences with intervening nouns of the opposite number, as a test of structural sensitivity. Networks given direct supervision handled most cases, but a plain language-modelling objective failed on the hard ones. The paper launched the targeted-evaluation style of probing neural networks for syntax, followed by [The emergence of number and syntax units in LSTM language models](https://md.research.elamur.ai/free/en/papers/lakretz-2019-emergence-number-syntax-units-lstm/), which traced agreement to a few individual units, and by [Does Syntax Need to Grow on Trees?](https://md.research.elamur.ai/free/en/papers/mccoy-2020-does-syntax-need-grow-trees/), which found that only tree-structured models reliably preferred hierarchical rules.

### The structural probe (2019)

[A Structural Probe for Finding Syntax in Word Representations](https://md.research.elamur.ai/free/en/papers/hewitt-2019-structural-probe-finding-syntax-word/) asked whether whole parse trees, not just isolated labels, can be read off contextual word vectors. A single learned linear transformation turned distances between vectors into distances in the dependency tree, and BERT's middle layers recovered trees far better than baselines. The method became a standard tool of interpretability research; its limits are questioned in [Mechanisms vs. Outcomes: Probing for Syntax Fails to Explain Performance on Targeted Syntactic Evaluations](https://md.research.elamur.ai/free/en/papers/agarwal-2025-mechanisms-vs-outcomes-probing-syntax/).

### Syntax-based statistical translation (2001)

[A Syntax-based Statistical Translation Model](https://md.research.elamur.ai/free/en/papers/yamada-2001-syntax-based-statistical-translation-model/) turned an English parse tree into a Japanese string through three stochastic operations per node — reordering children, inserting words and translating leaves — trained with EM. It was an early demonstration that a parser's tree could guide statistical translation, and it opened the line of tree-to-string models continued in [Statistical Syntax-Directed Translation with Extended Domain of Locality](https://md.research.elamur.ai/free/en/papers/huang-2006-statistical-syntax-directed-translation-extended/), [Syntax Augmented Machine Translation via Chart Parsing](https://md.research.elamur.ai/free/en/papers/zollmann-2006-syntax-augmented-machine-translation-chart/) and [Learning to Translate with Source and Target Syntax](https://md.research.elamur.ai/free/en/papers/chiang-2010-learning-translate-source-target-syntax/).

### Self-attentive constituency parsing (2018)

[Constituency Parsing with a Self-Attentive Encoder](https://md.research.elamur.ai/free/en/papers/kitaev-2018-constituency-parsing-self-attentive-encoder/) kept a span-based chart decoder but swapped the recurrent encoder for Transformer-style self-attention. Separating positional from content information improved accuracy, and with ELMo the parser set a new single-model record on the Penn Treebank. It built on [A Minimal Span-Based Neural Constituency Parser](https://md.research.elamur.ai/free/en/papers/stern-2017-minimal-span-based-neural-constituency/), which had shown that independent scoring of spans and labels was enough, and was extended to eleven languages with BERT in [Multilingual Constituency Parsing with Self-Attention and Pre-Training](https://md.research.elamur.ai/free/en/papers/kitaev-2019-multilingual-constituency-parsing-self-attention/). The released parser became one of the most widely used constituency tools.

### Wide-coverage CCG parsing (2004–2007)

Stephen Clark and James Curran's series made CCG practical on newspaper text. [Parsing the WSJ Using CCG and Log-Linear Models](https://md.research.elamur.ai/free/en/papers/clark-2004-parsing-wsj-ccg-log-linear/) trained discriminative models over packed charts on a cluster; [The Importance of Supertagging for Wide-Coverage CCG Parsing](https://md.research.elamur.ai/free/en/papers/clark-2004-importance-supertagging-wide-coverage-ccg/) showed that a supertagger which offers more categories only when the parser fails can speed parsing by an order of magnitude; and [Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models](https://md.research.elamur.ai/free/en/papers/clark-2007-wide-coverage-efficient-statistical-parsing/) gave the full journal account. The resulting C&C parser was for years the reference CCG system.

### CCGbank (2007)

[CCGbank](https://md.research.elamur.ai/free/en/papers/hockenmaier-2007-ccgbank-corpus-ccg-derivations-dependency/) by Julia Hockenmaier and Mark Steedman converted almost all of the Penn Treebank into CCG derivations, recovering long-range dependencies from extraction, control and coordination along the way. Nearly every statistical CCG parser since, from C&C to neural supertaggers, has been trained and evaluated on it.

### Synchronous TAG (1990)

[Synchronous Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/shieber-1990-synchronous-tree-adjoining-grammars/) by Stuart Shieber and Yves Schabes paired the elementary trees of two TAGs and rewrote them together. The second grammar might describe a logical form or another human language, giving a single mechanism for semantic interpretation and translation. Synchronous grammars later became central to syntax-based translation; see also [A Tree Transducer Model for Synchronous Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/maletti-2010-tree-transducer-model-synchronous-tree/).

### Mild context-sensitivity (1987)

[Characterizing Structural Descriptions Produced by Various Grammatical Formalisms](https://md.research.elamur.ai/free/en/papers/shanker-1987-characterizing-structural-descriptions-produced-various/) by K. Vijay-Shanker, David Weir and Aravind Joshi compared formalisms by the shape of their derivations. It found that CFG, head grammar, TAG and multicomponent TAG share context-free derivation structure and generalized them as Linear Context-Free Rewriting Systems, which have polynomial-time recognition. LCFRS remains the standard formal tool for discontinuous constituents.

### LFG on the Wall Street Journal (2002)

[Parsing the Wall Street Journal using a Lexical-Functional Grammar and Discriminative Estimation Techniques](https://md.research.elamur.ai/free/en/papers/riezler-2002-parsing-wall-street-journal-lexical/) combined a hand-written English LFG, the XLE parser and a log-linear disambiguation model. It reached full coverage through fragment and skimming fallbacks, and it trained the model by using the treebank's coarse brackets to filter the grammar's rich analyses. It showed that deep grammars and ordinary treebanks could be combined, and its companion [The Parallel Grammar Project](https://md.research.elamur.ai/free/en/papers/butt-2002-parallel-grammar-project/) described how LFG grammars for six languages were kept parallel at the level of functional structure.

### Probabilistic HPSG (2005–2008)

Yusuke Miyao and Jun'ichi Tsujii's [Probabilistic Disambiguation Models for Wide-Coverage HPSG Parsing](https://md.research.elamur.ai/free/en/papers/miyao-2005-probabilistic-disambiguation-models-wide-coverage/) and [Feature Forest Models for Probabilistic HPSG Parsing](https://md.research.elamur.ai/free/en/papers/miyao-2008-feature-forest-models-probabilistic-hpsg/) defined maximum entropy models over packed forests of feature structures and estimated them with a dynamic-programming analogue of the inside–outside algorithm. Their Enju parser brought HPSG analyses, including predicate–argument structure, to wide-coverage English text.

### Syntax in semantic role labelling (2008)

[The Importance of Syntactic Parsing and Inference in Semantic Role Labeling](https://md.research.elamur.ai/free/en/papers/punyakanok-2008-importance-syntactic-parsing-inference-semantic/) measured where full parses help a role labeller and found the main benefit in the early pruning of candidates. Its integer linear programming step, which enforces global constraints and merges systems built on different parsers, became a widely copied technique. A decade later [A Simple and Accurate Syntax-Agnostic Neural Model for Dependency-based Semantic Role Labeling](https://md.research.elamur.ai/free/en/papers/marcheggiani-2017-simple-accurate-syntax-agnostic-neural/) showed that a neural labeller could do well with no parse at all.

### Exact CCG parsing with A* (2014–2016)

[A* CCG Parsing with a Supertag-factored Model](https://md.research.elamur.ai/free/en/papers/lewis-2014-ccg-parsing-supertag-factored-model/) placed the entire statistical model in the supertagger and let a small set of combinators assemble the derivation, which made exact A* search fast. [LSTM CCG Parsing](https://md.research.elamur.ai/free/en/papers/lewis-2016-lstm-ccg-parsing/) replaced the tagger with a bidirectional LSTM and added tri-training, reaching new accuracy in and out of domain. Together they showed how far a strongly lexicalized formalism could be reduced to a tagging problem.

### Parsing morphologically rich languages (2013)

The [SPMRL 2013 shared task](https://md.research.elamur.ai/free/en/papers/seddah-2013-overview-spmrl-2013-shared-task/) compared constituency and dependency parsers on nine languages under gold, predicted and raw-lattice morphology. Its finding that ambiguous morphology, rather than the choice of framework, is the main obstacle set the agenda for joint models such as [Joint Transition-Based Models for Morpho-Syntactic Parsing](https://md.research.elamur.ai/free/en/papers/more-2019-joint-transition-based-models-morpho/).

## People

Authors with the most papers on this topic Stephen Clark leads with 7 papers; the 12 authors shown wrote 54 of the papers on this topic between them, counting each paper once per author.

**The numbers**

| Author | Papers here · their citations |
| --- | --- |
| [Stephen Clark](https://md.research.elamur.ai/free/en/authors/stephen-clark/) | 7 papers · 1,860 citations |
| [Kevin Knight](https://md.research.elamur.ai/free/en/authors/kevin-knight/) | 5 papers · 2,162 citations |
| [Luke S. Zettlemoyer](https://md.research.elamur.ai/free/en/authors/luke-zettlemoyer/) | 5 papers · 1,884 citations |
| [James R. Curran](https://md.research.elamur.ai/free/en/authors/james-r-curran/) | 5 papers · 1,246 citations |
| [Reut Tsarfaty](https://md.research.elamur.ai/free/en/authors/reut-tsarfaty/) | 5 papers · 492 citations |
| [Christopher D. Manning](https://md.research.elamur.ai/free/en/authors/christopher-d-manning/) | 4 papers · 4,322 citations |
| [Dan Klein](https://md.research.elamur.ai/free/en/authors/dan-klein/) | 4 papers · 1,473 citations |
| [Mark Steedman](https://md.research.elamur.ai/free/en/authors/mark-steedman/) | 4 papers · 1,343 citations |
| [Hai Zhao](https://md.research.elamur.ai/free/en/authors/hai-zhao/) | 4 papers · 475 citations |
| [Yue Zhang](https://md.research.elamur.ai/free/en/authors/yue-zhang/) | 4 papers · 438 citations |
| [Nathan Schneider](https://md.research.elamur.ai/free/en/authors/nathan-schneider/) | 4 papers · 36 citations |
| [Yoav Goldberg](https://md.research.elamur.ai/free/en/authors/yoav-goldberg/) | 3 papers · 1,645 citations |

- [Kevin Knight](https://md.research.elamur.ai/free/en/authors/kevin-knight/) (now at Texas Christian University) is the most prolific author on this topic. His papers here range from the syntax-based translation channel with Kenji Yamada and syntax-guided paraphrase lattices to [Does String-Based Neural MT Learn Source Syntax?](https://md.research.elamur.ai/free/en/papers/shi-2016-does-string-based-neural-mt/) and [Binarization of Synchronous Context-Free Grammars](https://md.research.elamur.ai/free/en/papers/huang-2009-binarization-synchronous-context-free-grammars/).
- [Stephen Clark](https://md.research.elamur.ai/free/en/authors/stephen-clark/) and [James R. Curran](https://md.research.elamur.ai/free/en/authors/james-r-curran/) built the C&C parser and its log-linear CCG models, wrote on supertagging and on evaluation against DepBank, and studied the parser's behaviour on Wikipedia text in [Evaluating a Statistical CCG Parser on Wikipedia](https://md.research.elamur.ai/free/en/papers/honnibal-2009-evaluating-statistical-ccg-parser-wikipedia/). Clark also appears with [Factorising AMR generation through syntax](https://md.research.elamur.ai/free/en/papers/cao-2019-factorising-amr-generation-through-syntax/).
- [Christopher D. Manning](https://md.research.elamur.ai/free/en/authors/christopher-d-manning/) (Stanford University) co-authored the typed-dependency converter with [Marie-Catherine de Marneffe](https://md.research.elamur.ai/free/en/authors/marie-catherine-de-marneffe/), the structural probe with [John Hewitt](https://md.research.elamur.ai/free/en/authors/john-hewitt/), and the Redwoods treebank paper.
- [Dan Klein](https://md.research.elamur.ai/free/en/authors/dan-klein/) (University of California, Berkeley) led the line of span-based neural constituency parsers with [Nikita Kitaev](https://md.research.elamur.ai/free/en/authors/nikita-kitaev/), and studied their behaviour outside the newspaper domain in [Cross-Domain Generalization of Neural Constituency Parsers](https://md.research.elamur.ai/free/en/papers/fried-2019-cross-domain-generalization-neural-constituency/).
- [Mark Steedman](https://md.research.elamur.ai/free/en/authors/mark-steedman/) (University of Edinburgh and Hong Kong University of Science and Technology), a principal architect of CCG, is co-author of CCGbank, the A* CCG parser with [Mike Lewis](https://md.research.elamur.ai/free/en/authors/mike-lewis/), and CCG grammar induction for semantic parsing in [Lexical Generalization in CCG Grammar Induction for Semantic Parsing](https://md.research.elamur.ai/free/en/papers/kwiatkowski-2011-lexical-generalization-ccg-grammar-induction/), with [Luke Zettlemoyer](https://md.research.elamur.ai/free/en/authors/luke-zettlemoyer/).
- [Julia Hockenmaier](https://md.research.elamur.ai/free/en/authors/julia-hockenmaier/) created CCGbank, the corpus on which statistical CCG parsing rests.
- [Aravind K. Joshi](https://md.research.elamur.ai/free/en/authors/aravind-joshi/), the founder of tree-adjoining grammar, appears here with the 1987 comparison of formalisms (with [K. Vijay-Shanker](https://md.research.elamur.ai/free/en/authors/k-vijay-shanker/) and [David J. Weir](https://md.research.elamur.ai/free/en/authors/david-weir/)) and with tree-to-string translation alongside [Liang Huang](https://md.research.elamur.ai/free/en/authors/liang-huang/) (Oregon State University).
- [Stuart M. Shieber](https://md.research.elamur.ai/free/en/authors/stuart-m-shieber/) and [Yves Schabes](https://md.research.elamur.ai/free/en/authors/yves-schabes/) introduced synchronous TAG.
- [Reut Tsarfaty](https://md.research.elamur.ai/free/en/authors/reut-tsarfaty/) (Bar-Ilan University) works on parsing morphologically rich languages, especially Hebrew: agreement in constituency parsing with [Khalil Sima'an](https://md.research.elamur.ai/free/en/authors/khalil-sima-an/), the SPMRL shared task with [Djamé Seddah](https://md.research.elamur.ai/free/en/authors/djame-seddah/), and joint morpho-syntactic parsing.
- [Tal Linzen](https://md.research.elamur.ai/free/en/authors/tal-linzen/) (New York University) and [Yoav Goldberg](https://md.research.elamur.ai/free/en/authors/yoav-goldberg/) (Bar-Ilan University) wrote the agreement study that started targeted syntactic evaluation of neural networks; Linzen continued it on hierarchical inductive bias with [R. Thomas McCoy](https://md.research.elamur.ai/free/en/authors/r-thomas-mccoy/) and [Robert Frank](https://md.research.elamur.ai/free/en/authors/robert-frank/).
- [Tracy H. King](https://md.research.elamur.ai/free/en/authors/tracy-holloway-king/) (now at Adobe), with [Ronald M. Kaplan](https://md.research.elamur.ai/free/en/authors/ronald-m-kaplan/), [Mark Johnson](https://md.research.elamur.ai/free/en/authors/mark-johnson/) and [Miriam Butt](https://md.research.elamur.ai/free/en/authors/miriam-butt/), represents the LFG grammar-engineering tradition of ParGram and XLE.
- [Yusuke Miyao](https://md.research.elamur.ai/free/en/authors/yusuke-miyao/) (The University of Tokyo) and [Jun'ichi Tsujii](https://md.research.elamur.ai/free/en/authors/jun-ichi-tsujii/) developed feature forests and the probabilistic HPSG parser Enju; [Stephan Oepen](https://md.research.elamur.ai/free/en/authors/stephan-oepen/) and [Dan Flickinger](https://md.research.elamur.ai/free/en/authors/dan-flickinger/) built the Redwoods treebank.
- [David Chiang](https://md.research.elamur.ai/free/en/authors/david-chiang/) (University of Notre Dame) showed how source and target syntax can be added to hierarchical translation as soft features, and co-authored a syntax-aware neural translation model.
- [Jacobus A. Naudé](https://md.research.elamur.ai/free/en/authors/jacobus-a-naude/) (University of the Free State) and [Cynthia L. Miller-Naudé](https://md.research.elamur.ai/free/en/authors/cynthia-l-miller-naude/) study Biblical and Qumran Hebrew syntax, including verbless clauses read through the Masoretic accents and the negation of participles.
- [Mohammed Q. Shormani](https://md.research.elamur.ai/free/en/authors/mohammed-q-shormani/) (Ibb University) writes minimalist analyses of Yemeni Arabic, such as [The syntax of wh-agreement in Yemeni Ibbi Arabic](https://md.research.elamur.ai/free/en/papers/naji-2026-syntax-wh-agreement-yemeni-ibbi/).
- [Nathan Schneider](https://md.research.elamur.ai/free/en/authors/nathan-schneider/) (Georgetown University) connects syntax to meaning representations, from AMR alignment to UCCA conversion and constructional semantics in language models.
- [Aarne Ranta](https://md.research.elamur.ai/free/en/authors/aarne-ranta/) created Grammatical Framework, described here as an interlingua for translation.

## Institutions and places

The authors with pages here work most often at the [University of Edinburgh](https://md.research.elamur.ai/free/en/universities/university-of-edinburgh/), long a centre of CCG research, followed by [Google](https://md.research.elamur.ai/free/en/companies/google/), [Bar-Ilan University](https://md.research.elamur.ai/free/en/universities/bar-ilan-university/), [Stanford University](https://md.research.elamur.ai/free/en/universities/stanford-university/) and the [University of Washington](https://md.research.elamur.ai/free/en/universities/university-of-washington/). Other homes include [Charles University](https://md.research.elamur.ai/free/en/universities/charles-university/), whose Prague tradition of deep syntax appears in [Translation of "It" in a Deep Syntax Framework](https://md.research.elamur.ai/free/en/papers/novak-2013-translation-it-deep-syntax-framework/); [Georgetown University](https://md.research.elamur.ai/free/en/universities/georgetown-university/); [Google DeepMind](https://md.research.elamur.ai/free/en/companies/google-deepmind/); the [Hong Kong University of Science and Technology](https://md.research.elamur.ai/free/en/universities/hong-kong-university-of-science-and-technology/); the [University of California, Berkeley](https://md.research.elamur.ai/free/en/universities/university-of-california-berkeley/); the [University of Colorado Boulder](https://md.research.elamur.ai/free/en/universities/university-of-colorado-boulder/); the [Centre National de la Recherche Scientifique](https://md.research.elamur.ai/free/en/labs/centre-national-de-la-recherche-scientifique/); [New York University](https://md.research.elamur.ai/free/en/universities/new-york-university/); [The University of Sydney](https://md.research.elamur.ai/free/en/universities/the-university-of-sydney/); the [University of Amsterdam](https://md.research.elamur.ai/free/en/universities/university-of-amsterdam/); and, for the linguistic studies of Semitic languages, the [University of the Free State](https://md.research.elamur.ai/free/en/universities/university-of-the-free-state/) and [Ibb University](https://md.research.elamur.ai/free/en/universities/ibb-university/).

Where the authors on this topic work now Google is the current home of authors of 10 papers on this topic; counted are the authors with a page here, by their latest institution.

**The numbers**

| Institution | Papers |
| --- | --- |
| [Google](https://md.research.elamur.ai/free/en/companies/google/) | 10 papers |
| [University of Edinburgh](https://md.research.elamur.ai/free/en/universities/university-of-edinburgh/) | 10 papers |
| [University of Washington](https://md.research.elamur.ai/free/en/universities/university-of-washington/) | 9 papers |
| [Centre National de la Recherche Scientifique](https://md.research.elamur.ai/free/en/labs/centre-national-de-la-recherche-scientifique/) | 8 papers |
| [Freeman Hospital](https://ror.org/00cdwy346) | 7 papers |
| [Bar-Ilan University](https://md.research.elamur.ai/free/en/universities/bar-ilan-university/) | 6 papers |
| [Charles University](https://md.research.elamur.ai/free/en/universities/charles-university/) | 5 papers |
| [Meta](https://md.research.elamur.ai/free/en/companies/meta/) | 5 papers |
| [New York University](https://md.research.elamur.ai/free/en/universities/new-york-university/) | 5 papers |
| [Stanford University](https://md.research.elamur.ai/free/en/universities/stanford-university/) | 5 papers |

By country, the United States leads, followed by the United Kingdom and Germany; Israel, China, the Netherlands, Australia and several others follow at a distance.

Countries of the authors on this topic United States comes first: authors working there now wrote 69 papers on this topic.

**The numbers**

| Country | Papers |
| --- | --- |
| [United States](https://md.research.elamur.ai/free/en/countries/united-states/) | 69 papers |
| [United Kingdom](https://md.research.elamur.ai/free/en/countries/united-kingdom/) | 28 papers |
| [Germany](https://md.research.elamur.ai/free/en/countries/germany/) | 16 papers |
| [China](https://md.research.elamur.ai/free/en/countries/china/) | 12 papers |
| [France](https://md.research.elamur.ai/free/en/countries/france/) | 9 papers |
| [Israel](https://md.research.elamur.ai/free/en/countries/israel/) | 9 papers |
| [Netherlands](https://md.research.elamur.ai/free/en/countries/netherlands/) | 7 papers |
| [India](https://md.research.elamur.ai/free/en/countries/india/) | 6 papers |
| [Czechia](https://md.research.elamur.ai/free/en/countries/czechia/) | 5 papers |
| [South Africa](https://md.research.elamur.ai/free/en/countries/south-africa/) | 5 papers |

The journal Computational Linguistics carries the largest number of these papers, followed by the Transactions of the Association for Computational Linguistics, then reference grammars and the general-linguistics journal Glossa. The rest are spread over ACL, NAACL, EMNLP and COLING proceedings and specialised workshops on grammar engineering and treebanks.

Where the papers on this topic were published Annual Meeting of the Association for Computational Linguistics (ACL) carries the most, 28 of 162.

**The numbers**

| Venue | Papers |
| --- | --- |
| [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/) | 28 papers |
| [Computational Linguistics](https://md.research.elamur.ai/free/en/venues/cl/) | 12 papers |
| [Conference on Empirical Methods in Natural Language Processing (EMNLP)](https://md.research.elamur.ai/free/en/venues/emnlp/) | 10 papers |
| [Language Resources and Evaluation Conference (LREC)](https://md.research.elamur.ai/free/en/venues/lrec/) | 9 papers |
| [Conference of the North American Chapter of the Association for Computational Linguistics (NAACL)](https://md.research.elamur.ai/free/en/venues/naacl/) | 8 papers |
| [Conference on Computational Natural Language Learning (CoNLL)](https://md.research.elamur.ai/free/en/venues/conll/) | 7 papers |
| [International Conference on Computational Linguistics (COLING)](https://md.research.elamur.ai/free/en/venues/coling/) | 7 papers |
| [Transactions of the Association for Computational Linguistics (TACL)](https://md.research.elamur.ai/free/en/venues/tacl/) | 6 papers |
| grammar-book | 5 papers |
| [Glossa a journal of general linguistics](https://md.research.elamur.ai/free/en/venues/glossa-a-journal-of-general-linguistics/) | 4 papers |

## Open questions

- **How much explicit syntax do neural models need?** Results here point both ways: syntax-aware encoders such as [Graph Convolutional Encoders for Syntax-aware Neural Machine Translation](https://md.research.elamur.ai/free/en/papers/bastings-2017-graph-convolutional-encoders-syntax-aware/) and [Improved Neural Machine Translation with a Syntax-Aware Encoder and Decoder](https://md.research.elamur.ai/free/en/papers/chen-2017-improved-neural-machine-translation-syntax/) gain over flat baselines, [Scheduled Multi-Task Learning: From Syntax to Translation](https://md.research.elamur.ai/free/en/papers/kiperwasser-2018-scheduled-multi-task-learning-syntax/) finds gains mainly in low-resource settings, and syntax-agnostic models match syntax-based ones on role labelling. Large pretrained models make the question sharper rather than settling it.
- **What does probing actually show?** A probe can find tree structure in a representation without that structure driving the model's behaviour, a gap stressed by recent work on mechanisms versus outcomes.
- **Robustness of deep grammars.** Hand-built grammars give precise analyses but still need fallbacks on unrestricted text; [Abstract Syntax as Interlingua](https://md.research.elamur.ai/free/en/papers/ranta-2020-abstract-syntax-as-interlingua-scaling/) describes the wide-coverage layer as unfinished work.
- **Fair evaluation across formalisms.** Conversion losses can be mistaken for parser errors, so scores need a conversion ceiling or a framework-neutral metric.
- **Morphologically rich and low-resource languages.** Semitic and other morphologically complex languages remain hard to parse, and how much of the difficulty lies in the languages, the models or treebank consistency is unresolved.
- **Syntax and meaning.** How syntactic structure aligns with semantic graphs such as AMR, UCCA and UMR is still being mapped: see [A Structured Syntax-Semantics Interface for English-AMR Alignment](https://md.research.elamur.ai/free/en/papers/szubert-2018-structured-syntax-semantics-interface-english/) and [Syntax is the Key to Semantics](https://md.research.elamur.ai/free/en/papers/heinecke-2026-syntax-is-key-semantics-combining/).
- **Theory and computation.** Whether competing theoretical frameworks — minimalist, constraint-based, construction-based — can be compared on shared data remains open; [Grammatical theory: From transformational grammar to constraint-based approaches](https://md.research.elamur.ai/free/en/papers/muller-2023-grammatical-theory-from-transformational/) surveys the landscape.

## Neighbouring topics

Syntax underlies most other structural analysis of language, so this topic touches many others. The largest overlap is with [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) and [Hebrew & Aramaic](https://md.research.elamur.ai/free/en/topics/hebrew-aramaic/), through descriptive studies of Biblical Hebrew, Arabic and Syriac syntax such as [Studies in the Syntax of the Peshitta of 1 Kings](https://md.research.elamur.ai/free/en/papers/williams-1998-studies-syntax-peshitta-1-kings/) and parsing work on Modern Hebrew. [Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) shares the long line of syntax-based translation models. [Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/), [Treebanks](https://md.research.elamur.ai/free/en/topics/treebanks/) and [Dependency parsing](https://md.research.elamur.ai/free/en/topics/dependency-parsing/) cover annotation and conversion, for example [A Universal Dependencies Conversion Pipeline for a Penn-format Constituency Treebank](https://md.research.elamur.ai/free/en/papers/arnardottir-2020-universal-dependencies-conversion-pipeline-penn/). [Morphology](https://md.research.elamur.ai/free/en/topics/morphology/) meets syntax in morphologically rich languages and in [A Rich Morphological Tagger for English](https://md.research.elamur.ai/free/en/papers/kirov-2017-rich-morphological-tagger-english-exploring/).

On the meaning side, [Semantic parsing & meaning representation](https://md.research.elamur.ai/free/en/topics/semantic-parsing/) and [Semantic role labeling](https://md.research.elamur.ai/free/en/topics/semantic-role-labeling/) use syntactic structure as input. Probing studies link the topic to [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) and [Interpretability](https://md.research.elamur.ai/free/en/topics/interpretability/). Smaller overlaps include [Coreference resolution](https://md.research.elamur.ai/free/en/topics/coreference-resolution/) (for example [Analysis of Syntax-Based Pronoun Resolution Methods](https://md.research.elamur.ai/free/en/papers/tetreault-1999-analysis-syntax-based-pronoun-resolution/)), [Grammatical error correction & learner language](https://md.research.elamur.ai/free/en/topics/grammatical-error-correction/) (through [SynGEC](https://md.research.elamur.ai/free/en/papers/zhang-2022-syngec-syntax-enhanced-grammatical-error/)), [Text generation](https://md.research.elamur.ai/free/en/topics/text-generation/) (CCG surface realisation and controlled paraphrase), [Historical linguistics & language change](https://md.research.elamur.ai/free/en/topics/language-change/), [Idioms & multiword expressions](https://md.research.elamur.ai/free/en/topics/multiword-expressions/), [Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/), [Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) and [Discourse analysis](https://md.research.elamur.ai/free/en/topics/discourse-analysis/).

Neighbouring topics: papers shared English shares the most: 47 of the 162 papers here (29%) stand on both.

**The numbers**

| Topic | Papers shared |
| --- | --- |
| [English](https://md.research.elamur.ai/free/en/topics/english/) | 47 shared · 29% |
| [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) | 26 shared · 16% |
| [Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) | 23 shared · 14% |
| [Hebrew & Aramaic](https://md.research.elamur.ai/free/en/topics/hebrew-aramaic/) | 14 shared · 9% |
| [Morphology](https://md.research.elamur.ai/free/en/topics/morphology/) | 11 shared · 7% |
| [Semantic parsing & meaning representation](https://md.research.elamur.ai/free/en/topics/semantic-parsing/) | 10 shared · 6% |
| [Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/) | 10 shared · 6% |
| [Biblical Hebrew](https://md.research.elamur.ai/free/en/topics/biblical-hebrew/) | 9 shared · 6% |
| [Treebanks](https://md.research.elamur.ai/free/en/topics/treebanks/) | 9 shared · 6% |
| [Evaluation & benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) | 6 shared · 4% |

Most papers on this topic stand on the [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) shelf, and many also on [Natural language processing](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) and [Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/).

The shelves this topic stands on Language and Linguistics: 156 of the 162 papers here.

**The numbers**

| Shelf | Papers |
| --- | --- |
| [Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) | 156 papers |
| [Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) | 119 papers |
| [Developmental and Educational Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/developmental-and-educational-psychology/) | 10 papers |
| [Computational Theory and Mathematics](https://md.research.elamur.ai/free/en/shelves/computer-science/computational-theory-and-mathematics/) | 9 papers |
| [Religious studies](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/religious-studies/) | 7 papers |
| [Cognitive Neuroscience](https://md.research.elamur.ai/free/en/shelves/neuroscience/cognitive-neuroscience/) | 2 papers |
| [Education](https://md.research.elamur.ai/free/en/shelves/social-sciences/education/) | 2 papers |
| [Experimental and Cognitive Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/experimental-and-cognitive-psychology/) | 2 papers |
| [Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) | 2 papers |
| [Philosophy](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/philosophy/) | 2 papers |
| [Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) | 1 paper |
| [Human-Computer Interaction](https://md.research.elamur.ai/free/en/shelves/computer-science/human-computer-interaction/) | 1 paper |

## Key works in this library

- [Generating Typed Dependency Parses from Phrase Structure Parses](https://md.research.elamur.ai/free/en/papers/marneffe-2006-generating-typed-dependency-parses-phrase/) — the Stanford typed-dependency scheme and its converter.
- [Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://md.research.elamur.ai/free/en/papers/linzen-2016-assessing-ability-lstms-learn-syntax/) — agreement as a test of structure in neural networks.
- [A Structural Probe for Finding Syntax in Word Representations](https://md.research.elamur.ai/free/en/papers/hewitt-2019-structural-probe-finding-syntax-word/) — parse trees recovered from vector geometry.
- [A Syntax-based Statistical Translation Model](https://md.research.elamur.ai/free/en/papers/yamada-2001-syntax-based-statistical-translation-model/) — an early tree-to-string translation channel.
- [Constituency Parsing with a Self-Attentive Encoder](https://md.research.elamur.ai/free/en/papers/kitaev-2018-constituency-parsing-self-attentive-encoder/) — self-attention for span-based parsing.
- [Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models](https://md.research.elamur.ai/free/en/papers/clark-2007-wide-coverage-efficient-statistical-parsing/) — the full account of the C&C parser.
- [CCGbank](https://md.research.elamur.ai/free/en/papers/hockenmaier-2007-ccgbank-corpus-ccg-derivations-dependency/) — the Penn Treebank converted to CCG.
- [The Importance of Syntactic Parsing and Inference in Semantic Role Labeling](https://md.research.elamur.ai/free/en/papers/punyakanok-2008-importance-syntactic-parsing-inference-semantic/) — where parses help role labelling.
- [Two/Too Simple Adaptations of Word2Vec for Syntax Problems](https://md.research.elamur.ai/free/en/papers/ling-2015-two-too-simple-adaptations-word2vec/) — order-aware embeddings for tagging and parsing.
- [Synchronous Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/shieber-1990-synchronous-tree-adjoining-grammars/) — paired grammars for semantics and translation.
- [Parsing the Wall Street Journal using a Lexical-Functional Grammar and Discriminative Estimation Techniques](https://md.research.elamur.ai/free/en/papers/riezler-2002-parsing-wall-street-journal-lexical/) — a deep grammar trained on a treebank.
- [The Parallel Grammar Project](https://md.research.elamur.ai/free/en/papers/butt-2002-parallel-grammar-project/) — parallel LFG grammars across languages.
- [Characterizing Structural Descriptions Produced by Various Grammatical Formalisms](https://md.research.elamur.ai/free/en/papers/shanker-1987-characterizing-structural-descriptions-produced-various/) — the origin of LCFRS.
- [A Minimal Span-Based Neural Constituency Parser](https://md.research.elamur.ai/free/en/papers/stern-2017-minimal-span-based-neural-constituency/) — span scoring without a grammar.
- [Syntax-based Alignment of Multiple Translations](https://md.research.elamur.ai/free/en/papers/pang-2003-syntax-based-alignment-multiple-translations/) — paraphrase lattices from merged parse trees.
- [Feature Forest Models for Probabilistic HPSG Parsing](https://md.research.elamur.ai/free/en/papers/miyao-2008-feature-forest-models-probabilistic-hpsg/) — log-linear models over packed feature structures.
- [Fast and Accurate Shift-Reduce Constituent Parsing](https://md.research.elamur.ai/free/en/papers/zhu-2013-fast-accurate-shift-reduce-constituent/) — a fast transition-based constituency parser.
- [Overview of the SPMRL 2013 Shared Task](https://md.research.elamur.ai/free/en/papers/seddah-2013-overview-spmrl-2013-shared-task/) — parsing morphologically rich languages.
- [A* CCG Parsing with a Supertag-factored Model](https://md.research.elamur.ai/free/en/papers/lewis-2014-ccg-parsing-supertag-factored-model/) — exact CCG parsing driven by supertags.
- [Learning to Translate with Source and Target Syntax](https://md.research.elamur.ai/free/en/papers/chiang-2010-learning-translate-source-target-syntax/) — syntax as soft features in translation.
- [Syntax-Guided Controlled Generation of Paraphrases](https://md.research.elamur.ai/free/en/papers/kumar-2020-syntax-guided-controlled-generation-paraphrases/) — paraphrases shaped by an exemplar's parse.
- [A Lexicalized Tree Adjoining Grammar for English](https://md.research.elamur.ai/free/en/papers/group-1998-lexicalized-tree-adjoining-grammar-english/) — the XTAG grammar of English.
- [Towards broad coverage surface realization with CCG](https://md.research.elamur.ai/free/en/papers/white-2007-broad-coverage-surface-realization-ccg/) — generating English from CCG logical forms.
- [At the interface of syntax and prosody](https://md.research.elamur.ai/free/en/papers/naude-2018-interface-syntax-prosody-differentiating-left/) — Biblical Hebrew verbless clauses read through the accents.
- [Germanic syntax: A constraint-based view](https://md.research.elamur.ai/free/en/papers/muller-2023-germanic-syntax-constraint-based/) — an HPSG account of Germanic clause structure.

## Papers (162)

**1** [2](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/2/) [3](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/3/) [next →](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/2/)

1. [**Generating Typed Dependency Parses from Phrase Structure Parses**Marie-Catherine de Marneffe, Bill MacCartney, Christopher D. Manning*2006 · 2,886 citations*](https://md.research.elamur.ai/free/en/papers/marneffe-2006-generating-typed-dependency-parses-phrase/)
2. **[Assessing the Ability of LSTMs to Learn Syntax-Sensitive Dependencies](https://md.research.elamur.ai/free/en/papers/linzen-2016-assessing-ability-lstms-learn-syntax/) [Transactions of the Association for Computational Linguistics (TACL)](https://md.research.elamur.ai/free/en/venues/tacl/)**Tal Linzen, Emmanuel Dupoux, Yoav Goldberg*2016 · 1,243 citations*
3. **[A Structural Probe for Finding Syntax in Word Representations](https://md.research.elamur.ai/free/en/papers/hewitt-2019-structural-probe-finding-syntax-word/) [Conference of the North American Chapter of the Association for Computational Linguistics (NAACL)](https://md.research.elamur.ai/free/en/venues/naacl/)**John Hewitt, Christopher D. Manning*2019 · 1,232 citations*
4. [**Learning to Map Sentences to Logical Form: Structured Classification with Probabilistic Categorial Grammars arXiv:1207.1420**Luke S. Zettlemoyer, Michael Collins*2012 · 1,074 citations*](https://md.research.elamur.ai/free/en/papers/zettlemoyer-2012-learning-map-sentences-logical-form/)
5. [**A Syntax-based Statistical Translation Model**Kenji Yamada, Kevin Knight*2001 · 1,019 citations*](https://md.research.elamur.ai/free/en/papers/yamada-2001-syntax-based-statistical-translation-model/)
6. **[Linguistic Knowledge and Transferability of Contextual Representations](https://md.research.elamur.ai/free/en/papers/liu-2019-linguistic-knowledge-transferability-contextual-representations/) [Conference of the North American Chapter of the Association for Computational Linguistics (NAACL)](https://md.research.elamur.ai/free/en/venues/naacl/)**Nelson F. Liu, Matt Gardner, Yonatan Belinkov, et al.*2019 · 1,015 citations*
7. **[Constituency Parsing with a Self-Attentive Encoder](https://md.research.elamur.ai/free/en/papers/kitaev-2018-constituency-parsing-self-attentive-encoder/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**Nikita Kitaev, Dan Klein*2018 · 762 citations*
8. [**Wide-Coverage Efficient Statistical Parsing with CCG and Log-Linear Models**Stephen Clark, James R. Curran*2007 · 626 citations*](https://md.research.elamur.ai/free/en/papers/clark-2007-wide-coverage-efficient-statistical-parsing/)
9. [**CCGbank: A Corpus of CCG Derivations and Dependency Structures Extracted from the Penn Treebank**Julia Hockenmaier, Mark Steedman*2007 · 612 citations*](https://md.research.elamur.ai/free/en/papers/hockenmaier-2007-ccgbank-corpus-ccg-derivations-dependency/)
10. **[The Importance of Syntactic Parsing and Inference in Semantic Role Labeling](https://md.research.elamur.ai/free/en/papers/punyakanok-2008-importance-syntactic-parsing-inference-semantic/) [Computational Linguistics](https://md.research.elamur.ai/free/en/venues/cl/)**Vasin Punyakanok, Dan Roth, Wen-tau Yih*2008 · 538 citations*
11. **[Two/Too Simple Adaptations of Word2Vec for Syntax Problems](https://md.research.elamur.ai/free/en/papers/ling-2015-two-too-simple-adaptations-word2vec/) [Conference of the North American Chapter of the Association for Computational Linguistics (NAACL)](https://md.research.elamur.ai/free/en/venues/naacl/)**Wang Ling, Chris Dyer, Alan W. Black, et al.*2015 · 504 citations*
12. **[Does String-Based Neural MT Learn Source Syntax?](https://md.research.elamur.ai/free/en/papers/shi-2016-does-string-based-neural-mt/) [Conference on Empirical Methods in Natural Language Processing (EMNLP)](https://md.research.elamur.ai/free/en/venues/emnlp/)**Xing Shi, Inkit Padhi, Kevin Knight*2016 · 496 citations*
13. **[Syntax Augmented Machine Translation via Chart Parsing](https://md.research.elamur.ai/free/en/papers/zollmann-2006-syntax-augmented-machine-translation-chart/) [Conference on Machine Translation (WMT)](https://md.research.elamur.ai/free/en/venues/wmt/)**Andreas Zollmann, Ashish Venugopal*2006 · 474 citations*
14. **[Synchronous Tree-Adjoining Grammars](https://md.research.elamur.ai/free/en/papers/shieber-1990-synchronous-tree-adjoining-grammars/) [International Conference on Computational Linguistics (COLING)](https://md.research.elamur.ai/free/en/venues/coling/)**Stuart M. Shieber, Yves Schabes*1990 · 405 citations*
15. **[Parsing the Wall Street Journal using a Lexical-Functional Grammar and Discriminative Estimation Techniques](https://md.research.elamur.ai/free/en/papers/riezler-2002-parsing-wall-street-journal-lexical/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**Stefan Riezler, Tracy H. King, Ronald M. Kaplan, et al.*2002 · 396 citations*
16. [**The Parallel Grammar Project COLING-02: Grammar Engineering and Evaluation**Miriam Butt, Helge Dyvik, Tracy Holloway King, et al.*2002 · 392 citations*](https://md.research.elamur.ai/free/en/papers/butt-2002-parallel-grammar-project/)
17. **[Parsing the WSJ Using CCG and Log-Linear Models](https://md.research.elamur.ai/free/en/papers/clark-2004-parsing-wsj-ccg-log-linear/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**Stephen Clark, James R. Curran*2004 · 388 citations*
18. **[Multilingual Constituency Parsing with Self-Attention and Pre-Training](https://md.research.elamur.ai/free/en/papers/kitaev-2019-multilingual-constituency-parsing-self-attention/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**Nikita Kitaev, Steven Cao, Dan Klein*2019 · 356 citations*
19. **[Characterizing Structural Descriptions Produced by Various Grammatical Formalisms](https://md.research.elamur.ai/free/en/papers/shanker-1987-characterizing-structural-descriptions-produced-various/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**K. Vijay-Shanker, David J. Weir, Aravind K. Joshi*1987 · 354 citations*
20. **[Syntax-Aware Aspect Level Sentiment Classification with Graph Attention Networks](https://md.research.elamur.ai/free/en/papers/huang-2019-syntax-aware-aspect-level-sentiment/) [Conference on Empirical Methods in Natural Language Processing (EMNLP)](https://md.research.elamur.ai/free/en/venues/emnlp/)**Binxuan Huang, Kathleen Carley*2019 · 351 citations*
21. **[Linguistically Motivated Large-Scale NLP with C&C and Boxer](https://md.research.elamur.ai/free/en/papers/curran-2007-linguistically-motivated-large-scale-nlp/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**James Curran, Stephen Clark, Johan Bos*2007 · 326 citations*
22. **[A Minimal Span-Based Neural Constituency Parser](https://md.research.elamur.ai/free/en/papers/stern-2017-minimal-span-based-neural-constituency/) [Annual Meeting of the Association for Computational Linguistics (ACL)](https://md.research.elamur.ai/free/en/venues/acl/)**Mitchell Stern, Jacob Andreas, Dan Klein*2017 · 309 citations*
23. [**Syntax-based Alignment of Multiple Translations: Extracting Paraphrases and Generating New Sentences**Bo Pang, Kevin Knight, Daniel Marcu*2003 · 289 citations*](https://md.research.elamur.ai/free/en/papers/pang-2003-syntax-based-alignment-multiple-translations/)
24. **[Lexical Generalization in CCG Grammar Induction for Semantic Parsing](https://md.research.elamur.ai/free/en/papers/kwiatkowski-2011-lexical-generalization-ccg-grammar-induction/) [Conference on Empirical Methods in Natural Language Processing (EMNLP)](https://md.research.elamur.ai/free/en/venues/emnlp/)**Tom Kwiatkowski, Luke Zettlemoyer, Sharon Goldwater, et al.*2011 · 284 citations*
25. [**Feature Forest Models for Probabilistic HPSG Parsing**Yusuke Miyao, Jun'ichi Tsujii*2008 · 279 citations*](https://md.research.elamur.ai/free/en/papers/miyao-2008-feature-forest-models-probabilistic-hpsg/)

**1** [2](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/2/) [3](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/3/) [next →](https://md.research.elamur.ai/free/en/topics/syntax-grammar/page/2/)

Updated 24 October 2026

- [← previous topic Summarization](https://md.research.elamur.ai/free/en/topics/summarization/)
- [next topic → Text generation](https://md.research.elamur.ai/free/en/topics/text-generation/)
- [↑ back to topics](https://md.research.elamur.ai/free/en/topics/)