---
title: "appos: appositional modifier — definition and key papers"
description: "appos (appositional modifier) is the Universal Dependencies relation linking a nominal to a second nominal placed directly after it that names…"
canonical: "https://research.elamur.ai/free/en/concepts/ud-appos/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/concepts/ud-appos/)

[home](https://md.research.elamur.ai/free/en/) / [concepts](https://md.research.elamur.ai/free/en/concepts/) / appos: appositional modifier

# 🌐 appos: appositional modifier

Universal Dependencies · 6 key papers · [all concepts](https://md.research.elamur.ai/free/en/concepts/)

Part of: [🗨️ Natural language processing](https://md.research.elamur.ai/free/en/concepts/natural-language-processing/) · [🌐 Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/)

appos (appositional modifier) is the Universal Dependencies relation linking a nominal to a second nominal placed directly after it that names, defines or describes the same entity, as in "Sam, my brother". The first nominal is the head.

Also: appositional modifier, apposition, appos

## What it is

Two noun phrases sometimes stand side by side and point to the same person or thing. In "Marie Curie, a physicist, won twice", the phrase "a physicist" adds nothing new to the event. It tells us more about who Marie Curie is. [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/) (UD) marks this pairing with the label appos.

UD is a cross-linguistic scheme in which words are joined by typed relations in a [dependency](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/) tree, with content words as heads ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). Version 2 keeps a closed set of 37 universal relation types, and languages may refine these only through subtypes ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). The label inventory goes back to the Universal Stanford Dependencies proposal ([Marneffe et al. 2014](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/)). It was then fixed, with 40 types, in the first UD guidelines ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)).

In the guidelines' sense, appos joins two nominals as follows:

- The head is the first nominal.
- The dependent is the head word of the second nominal, which usually follows directly, often set off by commas or brackets.
- The second phrase must be able to stand in for the first. It is a coreferent restatement, not a modifier that narrows the reference.
- Parenthesised abbreviations and spelled-out forms also count as appositions.

Several neighbouring relations have to be kept apart from appos:

- flat covers the parts of a single name, as in "Marie Curie". These are one name, not two phrases.
- compound covers noun–noun words such as "phone book".
- nmod is used when the second nominal is a separate entity related to the first, as in "the capital of France".
- conj and list are used when the phrases name different things.
- acl:relcl is used when the description is a clause.
- parataxis covers looser juxtaposition of clauses.

No widely shared subtype of appos exists across treebanks. Distinctions that a language wants to keep are usually made through features or neighbouring labels instead.

## Examples

| Sentence | Relation |
| --- | --- |
| Sam, my brother, arrived late. | appos(Sam → brother) |
| The World Health Organization (WHO) met today. | appos(Organization → WHO) |
| We visited Lyon, the city of lights. | appos(Lyon → city) |

In the third sentence, "of lights" attaches to "city" by nmod inside the appositive phrase. Only the head noun "city" carries the appos edge. Compare "the city of Lyon": here "Lyon" depends on "city" through nmod, because the phrase is a single reference with a prepositional complement, not two coreferent phrases.

## Across languages

Apposition is found in almost every written language. Treebanks differ in how often they use the label and where they draw its boundaries.

In the English UD treebanks of release 2.18, appos is a minor relation. Most of its edges link a common noun to another common noun or to a proper noun, followed by proper-noun–proper-noun pairs. French treebanks of the same release use appos about twice as often relative to their size. Their dominant pattern is a common noun followed by a proper noun, as in "le président, Macron". French also has a notable share of dependents tagged X, typically foreign or unanalysed material placed after a noun.

Part of this variation reflects style, since news and encyclopaedic text are rich in appositions. Part of it reflects annotation decisions about titles, names and labels. A head-plus-name sequence may be treated as appos in one treebank and as flat or compound in another. Questions of how a construction maps onto the relation set, rather than how each language marks it, are the subject of recent work on constructions and strategies in UD (Nivre 2025). Related work examines clauses without a canonical predicate (Nivre et al. 2026). Tutorials on UD principles treat these labelling boundaries as a recurrent source of difficulty for annotators (Zeman 2021).

## In treebanks and research

In the [treebanks](https://md.research.elamur.ai/free/en/topics/treebanks/) themselves, appos is one of the relations that consistency checks single out. Its boundary with flat, nmod and conj is easily drawn differently by different annotators. Systematic assessment of annotation consistency across UD corpora is one line of research on this problem ([Marneffe et al. 2017](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/)).

For parsers, appos is learned like any other label. UD data served as the basis of the CoNLL 2017 and 2018 shared tasks on parsing raw text in many languages ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). Those tasks produced a new generation of multilingual parsers ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)).

Appositions are short-range, but they are sensitive to punctuation and to the parser's choice of head. When a parser confuses appos with conj or list, one entity may be split into two, or two entities merged into one.

Downstream, the appos edge is attractive because it ties a name directly to a description of it. This is useful for:

- [coreference resolution](https://md.research.elamur.ai/free/en/concepts/coreference-resolution/), which can treat the two phrases as mentions of one entity;
- [named entity recognition](https://md.research.elamur.ai/free/en/concepts/named-entity-recognition/), which gains a type hint such as "a physicist";
- [information extraction](https://md.research.elamur.ai/free/en/concepts/information-extraction/), where patterns of the form "X, the Y of Z" yield relational facts.

The enhanced representation of English UD was designed to add edges and refine labels in the basic tree for exactly such natural-language-understanding applications ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). Its treatment of core arguments shows how UD weighs cross-lingual parallelism against language-specific detail (Zeman 2017), and the same balance shapes how appos is applied in each language.

How often the relation appos occurs: its share of all tokens Counted by code over the Universal Dependencies 2.18 treebanks of English and French (every token once, subtypes with their base relation).

**The numbers**

| Treebanks | Share |
| --- | --- |
| English treebanks | 0.41% of tokens (4,568 tokens) |
| French treebanks | 0.82% of tokens (8,965 tokens) |

What appos joins: the most frequent head tag → dependent tag The five most frequent in each language's treebanks, by tokens.

**The numbers**

| Pair | Tokens |
| --- | --- |
| English: NOUN → NOUN | 1,087 |
| English: NOUN → PROPN | 907 |
| English: PROPN → PROPN | 601 |
| English: PROPN → NOUN | 517 |
| English: NUM → NOUN | 206 |
| French: NOUN → PROPN | 3,339 |
| French: NOUN → NOUN | 2,389 |
| French: PROPN → NOUN | 1,605 |
| French: NOUN → X | 360 |
| French: PROPN → PROPN | 319 |

## Key papers

The works this page cites, the most cited first; citations as counted by OpenAlex plus the papers of this library that cite the work, read 2026-10-24.

- [Universal Dependencies v1: A Multilingual Treebank Collection](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/) — Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter et al. · 2016 · 1,741 citations
- [Universal Stanford dependencies: A cross-linguistic typology](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/) — Marie-Catherine de Marneffe, Timothy Dozat, Natalia Silveira et al. · 2014 · 627 citations
- [Universal Dependencies](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/) — Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre et al. · 2021 · 480 citations
- [Enhanced English Universal Dependencies: An Improved Representation for Natural Language Understanding Tasks](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/) — Sebastian Schuster, Christopher D. Manning · 2016 · 336 citations
- [Universal Dependencies v2: An Evergrowing Multilingual Treebank Collection](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/) — Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter et al. · 2020 · 325 citations
- [Assessing the Annotation Consistency of the Universal Dependencies Corpora](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/) — Marie-Catherine de Marneffe, Matias Grioni, Jenna Kanerva et al. · 2017 · 30 citations

## Related concepts

- [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/) — Universal Dependencies is a framework and an open collection of treebanks that annotates the grammar of many languages with one shared…
- [Dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/) — Dependency parsing is the task of recovering a sentence's syntactic structure as directed, labelled links between its words. Each word is…

## More papers

Topics: [🌐 Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/) · [🌳 Dependency parsing](https://md.research.elamur.ai/free/en/topics/dependency-parsing/)

Shelves: [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/)

- [← previous concept amod: adjectival modifier](https://md.research.elamur.ai/free/en/concepts/ud-amod/)
- [next concept → Artificial intelligence](https://md.research.elamur.ai/free/en/concepts/artificial-intelligence/)
- [↑ back to concepts](https://md.research.elamur.ai/free/en/concepts/)