🌐 mark: marker

Universal Dependencies · 6 key papers · all concepts

Part of: 🗨️ Natural language processing · 🌐 Universal Dependencies

mark (marker) is the Universal Dependencies relation that attaches a function word introducing a subordinate clause to that clause's predicate. Typical markers are subordinating conjunctions such as "because", "that" or "if", and the infinitival particle "to".

Also: marker, mark relation, subordinator attachment

What it is

In Universal Dependencies (UD), every word of a sentence hangs from another word, and each link carries a typed label. The relation mark covers one specific job. It links the small word that turns a clause into a dependent of something else, such as "if" in "if it rains", to the main predicate of that clause.

The direction of the arrow follows UD's central design choice. Content words are heads, and function words such as auxiliaries, adpositions and copulas depend on them (Marneffe et al. 2021). The head of mark is therefore the predicate of the subordinate clause, usually a verb, though an adjective or noun can head the clause in copular sentences. The dependent is the subordinator itself. The clause as a whole then attaches upward through a clausal relation such as advcl, ccomp, csubj or acl.

UD relations grew out of the universal Stanford dependencies (Marneffe et al. 2014). The present inventory has 37 universal relation types, which languages may refine with subtypes but not extend with new basic categories (Nivre et al. 2020).

Neighbouring relations

• case: an adposition or case particle on a nominal ("in the house"). mark is reserved for words introducing clauses.
• cc: a coordinating conjunction ("and", "or") that links conjuncts rather than subordinating them.
• aux and cop: verbal function words such as auxiliaries and copulas, not clause-introducers.
• advmod: relative and interrogative adverbs such as "when" or "where" play a role inside the clause and are usually advmod, not mark.

Subtypes

Subtypes of mark are few and language-specific. Examples include mark:rel for relativizing particles and mark:adv for adverbializing particles, as in some Chinese treebanks. Most treebanks use bare mark.

Examples

The arrow notation reads head → dependent.

In the third sentence the marker attaches to "tired", not to "was". This follows the content-head principle, under which the copula is itself a dependent of the predicate adjective.

Across languages

The relation is universal, but what fills the dependent slot varies.

In English, the most frequent pair in UD 2.18 is a verb heading the infinitival particle "to", tagged PART. Verb → subordinating conjunction comes a close second, with smaller shares of adjectives and nouns heading the marker.

French shows a different profile. Verb → SCONJ ("que", "si", "quand") leads. Verb → ADP comes next, because prepositions introducing infinitives are treated as markers: in "Il a décidé de partir", the edge runs partir → de. Some adverbs are also attached as markers.

In both languages a noun can head a mark dependent, typically in clauses complementing or modifying a noun. These differences reflect UD's aim of describing typologically diverse languages within one scheme (Marneffe et al. 2021). Different languages use different strategies for subordination: separate words, clitics or affixes. Only free subordinating words receive mark, since the word, not the morpheme, is the unit of annotation (Nivre et al. 2016). How UD separates constructions from the strategies that express them is discussed by Nivre (Nivre 2025).

In treebanks and research

Origins and motivation

Before UD, schemes disagreed so much that, on parallel Swedish, Danish and English sentences, no dependency matched across all three schemes at once (Nivre et al. 2016). Attaching subordinators uniformly to the clause predicate is one of the choices that makes clause structure parallel across languages. It also keeps predicate–argument links direct for downstream use (Marneffe et al. 2021).

Enhanced representations

Schuster and Manning's enhanced English UD appends the marker or preposition to a relation label in the enhanced graph, as in nmod:after (Schuster & Manning 2016). The lexical content of the subordinator then sits on the clausal edge itself. This helps information extraction systems that read off relations between content words.

Parsing and consistency

UD treebanks were the basis of the CoNLL 2017 and 2018 shared tasks on parsing raw text in many languages (Marneffe et al. 2021). In such dependency parsing work, mark is generally an easy, high-frequency relation. Errors concentrate where part-of-speech tagging confuses a subordinator with an adposition or adverb, which shifts the choice between mark, case and advmod.

The same boundary is a known source of cross-treebank variation. Tools for checking annotation consistency across UD corpora target exactly such divergences (Marneffe et al. 2017). Tutorials on UD principles and tools cover them as well (Zeman 2021).

Further readings are collected under Universal Dependencies, treebanks, English and French.

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.

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-mark/