---
title: "nsubj: nominal subject — definition and key papers"
description: "The nominal subject (nsubj) relation is the Universal Dependencies (UD) relation linking a clause's predicate to a nominal that is the clause's…"
canonical: "https://research.elamur.ai/free/en/concepts/ud-nsubj/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/concepts/ud-nsubj/)

[home](https://md.research.elamur.ai/free/en/) / [concepts](https://md.research.elamur.ai/free/en/concepts/) / nsubj: nominal subject

# 🌐 nsubj: nominal subject

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/)

The nominal subject (nsubj) relation is the Universal Dependencies (UD) relation linking a clause's predicate to a nominal that is the clause's syntactic subject. In active clauses that subject is typically the agent-like argument. The dependent is typically a noun, pronoun or proper noun, or another word heading a nominal phrase; clausal subjects take csubj instead.

Also: nominal subject, subject relation

## What it is

In the basic tree of [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/), every word of a sentence hangs from exactly one head through a typed relation. nsubj names one such link: the nominal that serves as subject of a clause. It is one of the core argument relations, which UD separates from other dependents ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). The label descends from the universal Stanford dependencies, which were proposed as a cross-linguistic typology of relations ([Marneffe et al. 2014](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/)).

The head is the predicate of the clause. UD makes content words heads and lets auxiliaries and copulas depend on them ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). As a result, the head is not always a verb. In "The soup is cold" the subject attaches to *cold*, and *is* hangs from *cold* as a copula. The dependent is the head word of the subject phrase.

UD encodes the type of the dependent in the label. Phrasal and clausal subjects therefore get different relations, nsubj versus csubj ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). Its main neighbours among the basic universal relations are:

| Relation | Dependent | Contrast with nsubj |
| --- | --- | --- |
| csubj | clause | the subject is a clause, not a nominal |
| obj | nominal | the object rather than the subject |
| obl | nominal | an oblique, e.g. the agent phrase of a passive |
| expl | expletive | a dummy subject such as *it* or *there* |

UD v2 lets languages add subtypes to the universal relations but not new basic categories. nsubj:pass for passive subjects is a standard example ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). Schuster and Manning's enhanced English graphs add nsubj:xsubj. This extra edge links a controlled subordinate verb to the subject it shares with the main verb that governs it ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)).

## Examples

- "The dog chased the cat." gives nsubj(chased, dog). *cat* is obj.
- "She is a doctor." gives nsubj(doctor, She). The nominal predicate is the head.
- "The cat was chased by the dog." gives nsubj:pass(chased, cat). The *by*-phrase is obl.
- "Running is fun." uses csubj(fun, Running) if *running* is analysed as a clause. If the gerund is treated as a noun, the analysis is nsubj(fun, Running).

## Across languages

UD aims to annotate typologically diverse languages in one scheme ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). The UD overview article gives a dedicated account of how different languages encode core arguments ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). Core arguments in UD are also the topic of a study by Zeman (Zeman 2017).

Languages differ in how nsubj surfaces:

- In pro-drop languages such as Spanish or Italian, the subject is often absent, so many clauses have no nsubj in the basic tree.
- In ergative languages, the guidelines assign nsubj to both the intransitive subject and the transitive agent, regardless of case marking.
- In clauses without a verb, the choice of head is not straightforward. Recent work treats nonprototypical predication and nonpredicational clauses explicitly (Nivre et al. 2026). A separate paper by Nivre frames such choices in terms of constructions and strategies (Nivre 2025).

## In treebanks and research

nsubj is annotated in every UD [treebank](https://md.research.elamur.ai/free/en/topics/treebanks/). A basic dependency tree is mandatory for each one, and an enhanced graph is optional ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). UD grew out of efforts to replace incompatible annotation schemes with one harmonised standard. Before UD, three parallel Swedish, Danish and English sentences were each annotated in their own language's scheme. No two of these schemes agreed on more than 40% of the dependencies, and not one dependency was common to all three ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)).

For [dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/), consistent subject labels make it possible to compare parsers across languages and to transfer models between them ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). The CoNLL shared tasks of 2017 and 2018, which parsed raw text in many languages, were built on UD ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)).

The basic tree can miss a direct link between a content word and its real subject, for instance the subject of a controlled verb. Schuster and Manning added this link to English enhanced graphs and released converters with Stanford CoreNLP ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). They aimed these graphs at shallow understanding applications. These include [information extraction](https://md.research.elamur.ai/free/en/topics/information-extraction/) tasks such as open-domain relation extraction, as well as biomedical text mining and unsupervised semantic parsing ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)).

How consistently corpora apply relations such as nsubj has itself been audited across UD ([Marneffe et al. 2017](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/)).

How often the relation nsubj 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 | 9.47% of tokens (106,678 tokens) |
| French treebanks | 6.87% of tokens (74,872 tokens) |

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

**The numbers**

| Pair | Tokens |
| --- | --- |
| English: VERB → PRON | 54,824 |
| English: VERB → NOUN | 16,640 |
| English: NOUN → PRON | 6,659 |
| English: VERB → PROPN | 5,998 |
| English: ADJ → PRON | 5,324 |
| French: VERB → PRON | 35,994 |
| French: VERB → NOUN | 20,379 |
| French: VERB → PROPN | 6,579 |
| French: NOUN → PRON | 2,470 |
| French: ADJ → NOUN | 2,405 |

## 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 NOUN: noun](https://md.research.elamur.ai/free/en/concepts/upos-noun/)
- [next concept → NUM: numeral](https://md.research.elamur.ai/free/en/concepts/upos-num/)
- [↑ back to concepts](https://md.research.elamur.ai/free/en/concepts/)