--- title: "SYM: symbol — definition and key papers — Elamur Research" description: "SYM (symbol) is the universal part-of-speech tag in Universal Dependencies for word-like entities that differ from ordinary words in form…" canonical: "https://research.elamur.ai/free/en/concepts/upos-sym/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/concepts/upos-sym/) [home](https://md.research.elamur.ai/free/en/) / [concepts](https://md.research.elamur.ai/free/en/concepts/) / SYM: symbol # 🏷️ SYM: symbol Universal Dependencies · 7 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/) SYM (symbol) is the universal part-of-speech tag in Universal Dependencies for word-like entities that differ from ordinary words in form, function or both, such as currency signs, mathematical operators, the percent sign and emoji. Also: symbol, UPOS SYM ## What it is In [Universal Dependencies](https://md.research.elamur.ai/free/en/concepts/universal-dependencies/), every syntactic word gets one tag from a closed list of 17 universal part-of-speech categories ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). SYM is the tag for symbols. These are written signs that work like words: they can be read aloud ("percent", "plus", "dollar") and can fill a slot in a phrase. They are not spelled with the ordinary letters of the language. A symbol differs from punctuation. A punctuation mark structures or delimits text. A symbol carries content of its own. The guidelines treat SYM as part of the morphological layer of [part-of-speech tagging](https://md.research.elamur.ai/free/en/concepts/part-of-speech-tagging/), together with lemmas and features ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). ### Borderline cases - SYM vs PUNCT. A hyphen or slash that joins or separates material is PUNCT. The same sign is SYM when it means "minus", "per" or "and/or". - SYM vs NUM. Digits and numerals are NUM. A sign next to them, such as %, $ or °, is SYM. - SYM vs NOUN. Spelled-out units ("dollars") and unit abbreviations ("km") are usually NOUN. The sign "$" is SYM. - SYM vs X. X is reserved for material that cannot be analysed in the language, such as foreign words or fragments. Symbols are analysable, so they get SYM. - Emoji and emoticons ("☺", ":)") are SYM, even when they work like interjections. ### Features and relations Symbols rarely inflect, so the FEATS column of a SYM token is normally empty. Treebank-specific marks such as Typo can still appear, since UD v2 introduced that feature ([Nivre et al. 2020](https://md.research.elamur.ai/free/en/papers/nivre-2020-universal-dependencies-v2-evergrowing-multilingual/)). The dependency relation depends on what the symbol does in the sentence. An operator such as "+" or "/" between conjuncts is attached as cc. A sign standing in for a preposition ("@" for "at") is attached as case. An emoticon is attached as discourse. A percent or currency sign that heads a quantity is attached as obl or nmod, with the number depending on it. ## Examples - English: "Prices rose 5 **%** last year." The sign % is SYM, and 5 depends on it as a numeral modifier. - English: "Meet me **@** the station **:)**". Here "@" is SYM attached as case, and the smiley is SYM attached as discourse. - English: "salt **&** pepper", "x **=** 2 **+** 3". These are signs read as words. - French: "une hausse de 3 **%**", "10 **€**". The symbols head the measure phrase. - German: "**§** 5 des Gesetzes". The section sign is SYM. ## Across languages The inventory of signs is largely shared across scripts. Their syntax follows the host language, because a symbol takes the position of the word it is read as. Treebanks differ in two main ways: - whether "$5" or "5%" is split into separate tokens during [tokenization](https://md.research.elamur.ai/free/en/concepts/tokenization/); - whether a sign used as a conjunction is tagged SYM or CCONJ. The relations SYM tokens take differ between languages: | Language | Dominant relations of SYM | | --- | --- | | [English](https://md.research.elamur.ai/free/en/topics/english/) | cc and case lead, followed by root, discourse and obl | | [French](https://md.research.elamur.ai/free/en/topics/french/) | nmod and obl lead, followed by cc, case and appos | The English pattern reflects web and social-media text, with its "/", "&", "@" and emoticons. The French pattern reflects measure phrases headed by % or currency signs. In both languages symbols make up only a tiny share of all tokens. ## In treebanks and research The twelve-category universal tagset of Petrov, Das and McDonald had no symbol class. Its categories were derived pragmatically from existing treebank guidelines, and stray signs fell under "." (punctuation) or X ([Petrov et al. 2012](https://md.research.elamur.ai/free/en/papers/petrov-2012-universal-part-speech-tagset/)). UD v1 expanded that set to 17 tags, adding PROPN and AUX among others. In doing so it separated SYM from PUNCT ([Nivre et al. 2016](https://md.research.elamur.ai/free/en/papers/nivre-2016-universal-dependencies-v1-multilingual-treebank/)). Its morphological features came from Interset, an interlingua of tagset drivers that decode and re-encode tags ([Zeman 2008](https://md.research.elamur.ai/free/en/papers/zeman-2008-reusable-tagset-conversion-tagset-drivers/)). The theoretical account of UD treats function-like elements as dependents of content words ([Marneffe et al. 2021](https://md.research.elamur.ai/free/en/papers/marneffe-2021-universal-dependencies/)). This explains why a symbol read as a preposition or conjunction attaches as case or cc. The principles and tooling behind these choices are also presented in tutorial form (Zeman 2021). Because the SYM/PUNCT/X boundary rests on judgement, it is a known source of disagreement between [treebanks](https://md.research.elamur.ai/free/en/topics/treebanks/) of the same language. Consistency checks on UD corpora find annotation errors of this kind ([Marneffe et al. 2017](https://md.research.elamur.ai/free/en/papers/marneffe-2017-assessing-annotation-consistency-universal-dependencies/)). A symmetric divergence over UPOS trigrams can flag treebank pairs whose tagging diverges, without relying on a trained tagger ([Aggarwal & Zeman 2020](https://md.research.elamur.ai/free/en/papers/aggarwal-2020-estimating-pos-annotation-consistency-different/)). For [dependency parsing](https://md.research.elamur.ai/free/en/concepts/dependency-parsing/), SYM is a small but noisy class. Errors on it mostly come from tokenization and from the cc versus punct decision. How often the tag SYM 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 | 0.16% of tokens (1,755 tokens) | | French treebanks | 0.09% of tokens (1,027 tokens) | The relations SYM words take most often The five most frequent in each language's treebanks, by tokens. **The numbers** | Pair | Tokens | | --- | --- | | English: cc | 341 | | English: case | 316 | | English: root | 170 | | English: discourse | 136 | | English: obl | 119 | | French: nmod | 172 | | French: obl | 157 | | French: cc | 134 | | French: case | 125 | | French: appos | 84 | ## 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 - [A Universal Part-of-Speech Tagset](https://md.research.elamur.ai/free/en/papers/petrov-2012-universal-part-speech-tagset/) — Slav Petrov, Dipanjan Das, Ryan McDonald · 2012 · 1,180 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 - [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 - [Reusable Tagset Conversion Using Tagset Drivers](https://md.research.elamur.ai/free/en/papers/zeman-2008-reusable-tagset-conversion-tagset-drivers/) — Daniel Zeman · 2008 · 181 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 - [Estimating POS Annotation Consistency of Different Treebanks in a Language](https://md.research.elamur.ai/free/en/papers/aggarwal-2020-estimating-pos-annotation-consistency-different/) — Akshay Aggarwal, Daniel Zeman · 2020 · 7 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… - [Part-of-speech tagging](https://md.research.elamur.ai/free/en/concepts/part-of-speech-tagging/) — Part-of-speech tagging is the task of assigning each token in a text a grammatical category, such as noun, verb, adjective or determiner… - [Morphology](https://md.research.elamur.ai/free/en/concepts/morphology/) — Morphology is the branch of linguistics that studies the internal structure of words. It covers how words are built from meaningful units… ## More papers Topics: [🔤 Part-of-speech tagging](https://md.research.elamur.ai/free/en/topics/part-of-speech-tagging/) · [🌐 Universal Dependencies](https://md.research.elamur.ai/free/en/topics/universal-dependencies/) Shelves: [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) - [← previous concept Subword tokenization](https://md.research.elamur.ai/free/en/concepts/subword-tokenization/) - [next concept → Text classification](https://md.research.elamur.ai/free/en/concepts/text-classification/) - [↑ back to concepts](https://md.research.elamur.ai/free/en/concepts/)