Negation in Universal Dependencies

Summary

The authors study how negation is encoded in Universal Dependencies, with semantic parsing from UD trees as the goal. A survey of treebanks shows that the two existing negation features, Polarity=Neg and PronType=Neg, are applied inconsistently across treebanks and even within one language, and often against the guidelines. In UD 2.15, Polarity=Neg appears in only 224 of 296 treebanks; English-EWT did not mark clausal negation at all. The authors argue that perfect consistency would still not solve the problem: these features sit on individual words and describe form, so the meaning of negation cannot be composed from them easily. They propose marking the predicate itself with two new features, Negated=+ and DoubleNegated=+, which say directly that the predicate is semantically under negation. For English they provide GREW rewrite rules that add these features to existing UD annotation automatically, covering plain clausal negation, negation affecting adjuncts, and negative quantifiers. The proposal fits the UniDive morphosyntactic parsing shared task, where function words pass their features up to their heads.

How Mova uses it

• not and n't: PART, advmod, Polarity=Neg: grounds the error rule en.errors.not-part (not/n't must be PART, XPOS RB, Polarity=Neg; the old neg relation no longer exists). The paper's observation that EWT lacked Polarity=Neg in UD 2.15 is set against Mova's own check that EWT 2.18 has it on all 2077 tokens. The seed reads this as a recent realignment that older models may not reflect.
• Double negation: don't know nothing: cited for English being a language without negative concord (unlike Spanish). This backs two warning rules, en.errors.double-negation-pronoun ("don't know nothing") and en.errors.double-negation-det ("didn't take no reason"). Both are written as tree patterns over the Polarity=Neg and PronType=Neg features.
• The proposed predicate-level Negated=+ feature is noted in the seed as background; Mova does not implement it.

Effectiveness in Mova

Not measured separately. The paper is one of two or three sources behind three small rules in the English error-detection rule set. Gold-data checks reported in the seeds: en.errors.not-part on EWT 2.18, 2077 checks with 0 violations; the double-negation rules fire 1 + 1 times on EWT 2.18, both in colloquial text.

👨‍🔬💥

Topics: Universal Dependencies · Treebanks

Cite

The paper itself, not this card. BibTeX from the library's record of the work.

@inproceedings{findlay2025negation,
  title = {{Negation in Universal Dependencies}},
  author = {Findlay, Jamie Yates and Haug, Dag Trygve Truslew},
  booktitle = {Proceedings of the Eighth Workshop on Universal Dependencies (UDW, SyntaxFest 2025)},
  year = {2025},
  url = {https://aclanthology.org/2025.udw-1.8/}
}

References

8 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Guillaume Bonfante, Bruno Guillaume, and Guy Perrier. 2018. Application of Graph Rewriting to Natural Language Processing. Number 1 in Logic, Linguistics and Computer Science set. ISTE & Wiley, London. find
2. Agata Savary, Sara Stymne, Verginica Barbu Mititelu, Nathan Schneider, Carlos Ramisch, and Joakim Nivre. 2023. PARSEME meets Universal Dependencies: getting on the same page in representing multiword expressions. Northern European Journal of Language Technology, 9(1). find
3. Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre, and Daniel Zeman. 2021. Universal Dependencies. Computational Linguistics, 47(2):255–308. find our card
4. Leonie Weissweiler, Nina Böbel, Kirian Guiller, Santiago Herrera, Wesley Samuel Scivetti, Arthur Lorenzi, Nurit Melnik, Archna Bhatia, Hinrich Schütze, Lori Levin, Amir Zeldes, Joakim Nivre, William Croft, and Nathan Schneider. 2024. UCxn: Typologically-informed annotation of constructions atop Un… find
5. Jamie Y. Findlay, Saeedeh Salimifar, Ahmet Yıldırım, and Dag T. T. Haug. 2023. Rule-based semantic interpretation for Universal Dependencies. In Proceedings of the Sixth Workshop on Universal Dependencies (UDW, GURT/SyntaxFest 2023), pages 47– 57, Washington, D.C. Association for Computational Lingu… find
6. Bruno Guillaume, Guillame Bonfante, Paul Masson, Mathieu Morey, and Guy Perrier. 2012. Grew : un outil de réécriture de graphes pour le TAL (Grew: a graph rewriting tool for NLP) [in French]. In Proceedings of the Joint Conference JEP-TALN-RECITAL 2012, volume 5: Software Demonstrations, pages 1–2… find
7. Hedde Zeijlstra. 2020. Negative quantifiers. In Viviane Déprez and M. Teresa Espinal, editor, The Oxford handbook of negation, pages 426–440. Oxford University Press. find
8. Daniel Zeman et al. 2024. Universal Dependencies 2.15. LINDAT/CLARIAH-CZ digital library at the Institute of Formal and Applied Linguistics (ÚFAL), Faculty of Mathematics and Physics, Charles University. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/findlay-2025-negation-universal-dependencies/