Universal Dependencies guidelines (v2): English-specific pages

What it is about

The UD guidelines are the annotation rules for the universal treebanks. Volume 2 holds the English-specific pages. The parts read were: relative clauses (the notable properties, with a chain of acl:relcl and xcomp arcs that gives unbounded dependency, and the note that the specifying/ascriptive distinction does not change the UD analysis); clefts and pseudo-clefts (a free relative as the copula's complement, with nsubj on the relative word); fixed expressions (for example "as well" and "as well as", with advmod on the head and fixed inside the expression); and personal names (a title such as "Mr." as nmod:desc, with a note that English treebanks tag such titles as PROPN). The text is a markdown export with examples in sdparse blocks. No counts are given; these are rules, not measurements.

Ideas

• I1. Put the guideline examples of relative clauses, pseudo-clefts and fixed expressions into a test file for en/src/parse.rs. Score the parser on them and report which arcs differ from the documented ones. Add a shuffled-word negative control so that a passing score cannot be an accident.
• I2. Build a lexicon of English fixed expressions (as well, as well as, in spite of and similar) and a rule that attaches the later words as fixed to the first, in en/src/dict.rs. Measure the fixed arc accuracy on EWT dev before and after, run twice.
• I3. Pseudo-clefts ("X is who we want to help") are a foregrounding pattern. Add a rule in prag/src/frames.rs that marks the subject of such a copular sentence as the foregrounded element, and log each case so the rule can be checked by hand.
• I4. Titles before names ("Mr.", "Professor") are tagged NOUN in some schemes and PROPN in English treebanks. Make the tag rule in en/src/tag.rs explicit and test it on the personal-name examples, so that the choice is recorded and not silent.

For Mova

The chapters are 22. Phrases: noun, adjective, adverb and prepositional phrases, 25. Relative clauses, 29. Ellipsis, word order and information structure and 19. The lexicon: vocabulary, multiword expressions and register. The code is en/src/parse.rs (arcs), en/src/dict.rs (fixed expressions), en/src/tag.rs (titles) and prag/src/frames.rs (clefts). The measure is the arc accuracy on the guideline examples and on EWT dev, run twice and compared with the negative control.

Topics: Universal Dependencies · Treebanks

Cite

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

@book{contributors2026universal,
  title = {{Universal Dependencies guidelines (v2): English-specific pages}},
  author = {contributors, Universal Dependencies},
  year = {2026},
  url = {https://universaldependencies.org/guidelines.html}
}

References

4 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. [Loos, Eugene E., et al. 2003. Glossary of linguistic terms: What is an auxiliary verb?](http://www-01.sil.org/linguistics/GlossaryOfLinguisticTerms/WhatIsAnAuxiliaryVerb.htm) - [Wikipedia](http://en.wikipedia.org/wiki/Auxiliary_verb) ↗
2. For _subordinating conjunctions,_ see [SCONJ](). ↗
3. English Web Treebank [LDC2012T13](https://catalog.ldc.upenn.edu/LDC2012T13). ↗
4. Justin Mott, Ann Bies, John Laury, and Colin Warner. 2012. [Bracketing Webtext: An Addendum to Penn Treebank II Guidelines](https://catalog.ldc.upenn.edu/docs/LDC2012T13/WebtextTBAnnotationGuidelines.pdf). MS, LDC. # Migration Guidelines (from SD to UD) ↗

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/contributors-2026-universal-dependencies-guidelines/