🌐 obj: object

Universal Dependencies · 6 key papers · all concepts

Part of: 🗨️ Natural language processing · 🌐 Universal Dependencies

obj (object) is the Universal Dependencies relation linking a predicate, usually a verb, to its second core nominal argument: in active clauses, typically the participant most directly affected by the event, such as "the letter" in "She wrote the letter."

Also: object, direct object, dobj (UD v1 name), second core argument

What it is

In Universal Dependencies (UD), every word of a sentence attaches to one head with a typed label. The label obj marks the object, roughly what school grammar calls the direct object. In "Kim reads books", the verb "reads" is the head and "books" is the dependent.

The guidelines treat obj as one of the core arguments. Core arguments are kept apart from oblique dependents, which express optional circumstances or less central roles (Nivre et al. 2016; Zeman 2017). In UD v1 the same relation was called dobj. That inventory paired a nominal label with a clausal one, so that dobj stood beside ccomp just as nsubj stood beside csubj (Nivre et al. 2016). Version 2 shortened the name to obj and kept it among the 37 universal relation types (Nivre et al. 2020).

Because UD makes content words heads, the head of obj is the lexical predicate, not an auxiliary (Marneffe et al. 2021). The dependent is a nominal: a noun, a pronoun, a proper name, or a word used as a noun.

The neighbouring relations are:

Subtypes are rare. Languages may add their own extensions, written with a colon (Nivre et al. 2020). Some treebanks use one for the noun of a light-verb construction, written obj:lvc.

Examples

• "The cat caught a mouse": obj(caught, mouse).
• "We saw her yesterday": obj(saw, her). The pronoun is the object, while "yesterday" attaches as advmod.
• "She gave the boy a book": obj(gave, book) and iobj(gave, boy). The recipient takes iobj.
• "She gave a book to the boy": obj(gave, book) and obl(gave, boy). The prepositional phrase is oblique, with "to" attached as case to "boy".
• "He said that it rained": the clause is ccomp, not obj.

Across languages

UD defines core arguments by how a language actually encodes them: case, agreement or word order (Marneffe et al. 2021; Zeman 2017). This lets one label apply across very different typologies.

• In English, objects are recognised mainly by position after the verb. In English treebanks, verb-to-noun objects dominate, followed by verb-to-pronoun objects.
• In French, clitic pronouns such as "le" or "la" stand before the verb but are still obj, so verb-to-pronoun pairs are frequent there too. A smaller pattern has an adverb as head. Presentatives such as "voici", tagged as adverbs, take an obj, as in "voici le livre".
• In languages with differential object marking, a marked object (Spanish "a", Hindi "ko") remains obj as long as it behaves as a core argument.
• In ergative languages, obj still labels the patient-like argument of a transitive verb. The identity of the relation does not depend on the case it takes (Zeman 2017).

Hard cases include objects of nonverbal predicates and argument-like nominals in constructions that are not prototypical predications. Recent work revisits how such strategies are annotated (Nivre 2025; Nivre et al. 2026).

In treebanks and research

The relation descends from the direct-object label of Stanford dependencies. That label was generalised in the cross-linguistic typology that preceded UD (Marneffe et al. 2014).

In the enhanced representation, an object shared by conjoined verbs is propagated to each conjunct. This adds direct edges useful for information extraction (Schuster & Manning 2016).

For dependency parsing, obj is among the labels whose accuracy is reported most often. The CoNLL 2017 and 2018 shared tasks evaluated parsers on UD data across many languages (Marneffe et al. 2021; Nivre et al. 2020). A common label makes cross-lingual transfer possible, for example training on one language and parsing another. It also offers a syntactic proxy for the patient role in semantic role labeling.

The boundary between obj and obl is not always drawn the same way across treebanks. Annotation-consistency studies therefore check how uniformly core arguments are labelled (Marneffe et al. 2017). Teaching material stresses the same distinction (Zeman 2021).

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-obj/