🌐 cop: copula
cop (copula) is the Universal Dependencies relation that attaches a copular function word, such as English "be", to a nonverbal predicate. The noun, adjective, pronoun or adverb is the head of the clause, and the copula is its dependent.
What it is
In Universal Dependencies (UD), a sentence such as "Anna is a doctor" gets an unexpected tree. The verb "is" is not the root. The noun "doctor" heads the clause, and "is" hangs from it under the label cop. This follows a general design choice: content words are heads and function words depend on them. Copulas are treated alongside auxiliaries and adpositions (Marneffe et al. 2021). The choice goes back to the cross-linguistic typology of Universal Stanford Dependencies (Marneffe et al. 2014). UD v1 kept it when it merged earlier schemes into one standard (Nivre et al. 2016).
In the guidelines' terms, a copula is a function word whose only job is to link a subject to a nonverbal predicate. That predicate may be a nominal, an adjective, an adverb or a prepositional phrase. The copula may also carry tense, mood or agreement.
The relation joins two words:
- The head is the predicate word (NOUN, ADJ, PRON, ADV, PROPN, or the noun inside a predicative phrase).
- The dependent is the copula, normally tagged AUX in part-of-speech tagging.
The subject then attaches to the predicate as nsubj, not to the copula. This keeps copular clauses parallel to languages that have no copula at all, which increases cross-lingual uniformity (Marneffe et al. 2021).
Neighbouring relations
cop must be told apart from several nearby relations:
- aux covers tense, aspect and voice auxiliaries attached to a verbal head.
- xcomp covers verbs like "become" or "seem", which UD v2 treats as full verbs taking a predicative complement. Only the pure linking verb counts as a copula.
- Existential "there is" and possessive "have" are ordinary verbal predicates.
Subtypes are permitted in UD in general (Nivre et al. 2020). No cop:subtype is in common cross-linguistic use, so the relation usually appears bare.
Examples
Three short English sentences, with the relation written as cop(head, dependent):
| Sentence | Head | Relation |
|---|---|---|
| The soup is cold. | cold (ADJ) | cop(cold, is) |
| My sister was the winner. | winner (NOUN) | cop(winner, was) |
| The keys are on the table. | table (NOUN) | cop(table, are); case(table, on) |
In the last sentence, the noun inside the prepositional phrase becomes the clausal head, and the preposition attaches to it as case. In "She is here", the adverb "here" is the head.
Across languages
Languages differ mainly in whether a copula word exists and when it is spelled out.
- Zero copula: Russian in the present tense, Hungarian in the third person, and Arabic nominal sentences (Semitic languages) often have no copula. The predicate simply heads the clause with no cop dependent.
- Overt copula: when a copula does appear, the tree differs from the copula-less version only by that one extra edge. This is the parallelism the design aims for (Marneffe et al. 2021).
- Pronominal copulas: Modern Hebrew uses third-person pronouns in a linking function. Treebanks may attach such words as cop even though they are not verbs.
- Two linking verbs: Spanish and Portuguese "ser" and "estar" are both annotated as copulas.
In the English and French UD treebanks, nouns and adjectives are by far the most frequent heads of cop. English then shows notable pronoun and adverb heads. French shows some verbal heads, typically infinitival or clausal predicates.
The boundary of "nonverbal predication" is not fixed. Recent work examines clauses whose predication is nonprototypical or which lack predication altogether (Nivre et al. 2026). It also asks how constructions relate to the strategies languages use to express them (Nivre 2025). The treatment of copular subjects connects to the wider question of how core arguments are encoded (Zeman 2017).
In treebanks and research
For dependency parsing, cop is a demanding relation in three ways:
- The parser must recognise a nonverbal word as the clause root.
- It must decide whether a form like "be" is a copula, an auxiliary or a main verb.
- It must reattach the subject and any clausal dependents to the predicate.
Annotators make similar errors. Consistency checks across UD treebanks target exactly such recurrent word–relation patterns (Marneffe et al. 2017). Tutorials on UD present the copula analysis as a standard case of its principles (Zeman 2021).
Downstream, the content-head analysis is meant to make predicate–argument structure transparent (Marneffe et al. 2021). In "Anna is a doctor", the subject edge links "Anna" directly to "doctor". Enhanced UD graphs build on such basic trees by adding edges and refining labels for applications such as relation extraction (Schuster & Manning 2016).
Critics note that copula-headed alternatives can be easier to learn in some languages. Treating the copula as a dependent is nonetheless a central part of the uniform scheme used in multilingual parsing and in the CoNLL shared tasks (Nivre et al. 2020).
What cop 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 → AUX | 9,795 |
| English: ADJ → AUX | 9,534 |
| English: PRON → AUX | 2,671 |
| English: ADV → AUX | 1,946 |
| English: PROPN → AUX | 988 |
| French: NOUN → AUX | 5,914 |
| French: ADJ → AUX | 5,206 |
| French: PRON → AUX | 910 |
| French: VERB → AUX | 701 |
| French: ADV → AUX | 257 |
Key papers
Related 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 — 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 · 🌳 Dependency parsing
Shelves: 🗣️ Language and Linguistics
This page in Markdown → https://md.research.elamur.ai/free/en/concepts/ud-cop/