--- title: "orphan: orphan — definition and key papers — Elamur Research" description: "`orphan` is the Universal Dependencies (UD) relation used in elliptical clauses whose predicate has been left out." canonical: "https://research.elamur.ai/free/en/concepts/ud-orphan/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/concepts/ud-orphan/) [home](https://md.research.elamur.ai/free/en/) / [concepts](https://md.research.elamur.ai/free/en/concepts/) / orphan: orphan # 🌐 orphan: orphan 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/) `orphan` is the Universal Dependencies (UD) relation used in elliptical clauses whose predicate has been left out. One surviving dependent is promoted to stand in for the missing predicate, and each other surviving dependent attaches to that promoted word with `orphan`. Also: remnant (the earlier UD v1 approach it replaced) ## What it is Speakers often leave a verb unsaid when it can be recovered from context: «Mary ordered tea and John coffee.» The second clause has no verb, but it still has a subject and an object. In [dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/), each clause of the tree still needs one word as its head. [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/) handles this by promotion. One leftover dependent takes the place of the missing predicate, and the other leftovers attach to it with `orphan`. In UD, content words are heads and function words depend on them, and every treebank must provide a basic tree ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/); [Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). Promotion keeps that tree intact when the content-word head is absent. ### Head and dependent The relation mostly occurs in gapping. In gapping, a clause repeats the structure of a preceding clause but omits its verb. The head of `orphan` is the promoted remnant, and in gapping that remnant is itself attached to the first clause's verb by `conj` (conjunct). The dependent is another remnant. Had the verb been present, it would have depended on that verb alongside the promoted word. The guidelines choose the word to promote by an obliqueness hierarchy. The remnant whose function ranks highest becomes the head: | Rank | Function of remnant | | --- | --- | | 1 | `nsubj` (nominal subject) | | 2 | `obj` (direct object) | | 3 | `iobj` (indirect object) | | 4 | `obl` (oblique nominal) | | 5 | `advmod` (adverbial modifier) | | 6 | `csubj` (clausal subject) | | 7 | `xcomp` (open clausal complement) | | 8 | `ccomp` (clausal complement) | | 9 | `advcl` (adverbial clause) | The hierarchy places core arguments above oblique dependents. How that distinction applies across UD languages is examined in (Zeman 2017). The label is reserved for cases where the original relation cannot be kept. If an auxiliary or copula survives, that word is promoted instead, and the ordinary labels stay in place. In «John has finished and Mary has too», the second *has* heads the clause. *Mary* stays `nsubj` and *too* attaches as `advmod`, so no `orphan` appears. ### Neighbouring relations - `conj` links the elliptical clause to its antecedent clause, while `orphan` works inside the elliptical clause. - `dep` (unspecified dependency) is the label for an unclear relation. `orphan` instead records a clear relation whose real head is missing. - `parataxis` (loosely juxtaposed clauses) and `list` (items of a list) join loosely related units, not remnants of one clause. ### History and enhanced representation The inventory of 37 universal relation types in v2 includes `orphan` ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). It took over from the v1 `remnant` analysis, which belonged to an inventory of 40 types ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). No subtype of `orphan` is defined universally. Languages may add subtypes to any relation ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)), but forms such as `orphan:…` are not in common use. The optional enhanced representation is a directed graph that adds implicit relations to the basic tree ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). In it, the missing verb can be restored as an empty node: an extra node that stands for the omitted word without matching any token. The remnants then receive their true labels, so `orphan` appears only in the basic tree. Enhanced structures give applications that need relations between content words a better input than the bare tree ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). UD empty nodes differ in detail from the copy nodes that the enhanced English proposal borrows for conjoined prepositions ([Schuster & Manning 2016](https://md.research.elamur.ai/free/en/papers/schuster-2016-enhanced-english-universal-dependencies-improved/)). ## Examples Each arc is written head → relation → dependent. 1. «Mary won gold and Peter bronze.» won → `conj` → Peter; Peter → `orphan` → bronze 2. «I go on Mondays and she on Fridays.» go → `conj` → she; she → `orphan` → Fridays 3. «He wants pizza, and his sister, probably, pasta.» wants → `conj` → sister; sister → `orphan` → probably; sister → `orphan` → pasta In the third example, two instances of `orphan` hang from the same promoted subject. That is the expected pattern when more than two dependents survive. ## Across languages `orphan` is rare in the treebanks examined, since genuine predicate ellipsis is uncommon in running text. In the [English](https://md.research.elamur.ai/free/en/topics/english/) data, the most frequent head–dependent pairs appear to be a noun on a noun, followed by proper-noun heads over nouns or proper nouns. This matches subject-plus-object gapping. Some pairs have a pronoun head over an adverb («Who spoke first, and who next?»). In [French](https://md.research.elamur.ai/free/en/topics/french/), noun-on-noun pairs dominate as well. Pairs with a noun or proper noun over a numeral are frequent, as in comparative or enumerative gapping such as «Paris en compte trois, Lyon deux». Languages with freer word order, rich case marking or frequent verbless clauses may promote different remnants. They also differ in whether a nominal predicate is analysed with `orphan` or as an ordinary copula-less clause. Such boundary cases belong to the wider question of nonverbal and nonprototypical predication in UD (Nivre et al. 2026). They can also be framed through the distinction between constructions and the strategies languages use to express them (Nivre 2025). ## In treebanks and research UD aims to annotate constructions consistently across very different languages, so that predicate–argument structure stays visible to downstream tools ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). `orphan` itself arrived with UD v2. The `remnant` analysis it replaced traces back to the Universal Stanford Dependencies typology ([Marneffe et al. 2014](https://md.research.elamur.ai/free/en/papers/marneffe-2014-universal-stanford-dependencies-cross-linguistic/)). Because `orphan` is so infrequent, its annotation is likely to be less uniform across [treebanks](https://md.research.elamur.ai/free/en/topics/treebanks/) than that of common labels. General consistency checks of UD corpora can expose such inconsistencies ([Marneffe et al. 2017](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/)). For [parsing](https://md.research.elamur.ai/free/en/concepts/parsing/), the relation is a known hard case. A parser must notice that a verb is missing and pick the right word to promote. It must then distinguish `orphan` from `conj`, `obj` or `dep`, all with very little training data. UD supplied the data for the CoNLL 2017 and CoNLL 2018 shared tasks, which evaluated multilingual parsing starting from raw text ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/); [Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). How often the relation orphan 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 | 0.02% of tokens (206 tokens) | | French treebanks | 0.05% of tokens (505 tokens) | What orphan 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 | 37 | | English: PROPN → NOUN | 18 | | English: PROPN → PROPN | 17 | | English: PRON → ADV | 14 | | English: NOUN → ADV | 13 | | French: NOUN → NOUN | 118 | | French: NOUN → PROPN | 35 | | French: PRON → NOUN | 33 | | French: NOUN → NUM | 28 | | French: NOUN → ADV | 26 | ## 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 Ontology engineering](https://md.research.elamur.ai/free/en/concepts/ontology-engineering/) - [next concept → parataxis: parataxis](https://md.research.elamur.ai/free/en/concepts/ud-parataxis/) - [↑ back to concepts](https://md.research.elamur.ai/free/en/concepts/)