Manning's Law
{{neologism|date=March 2019}}
Manning's Law describes the combination of principles that need to be balanced in the design and growth of universal linguistic dependencies. These dependencies are used to describe and model syntactic relations, for all languages.{{cite journal|last1=de Marneffe|first1=Marie-Catherine|last2=Manning|first2=Christopher D.|last3=Nivre|first3=Joakim|last4=Zeman|first4=Daniel|title=Universal Dependencies|journal=Computational Linguistics|date=13 July 2021|volume=47|number=2|pages=255–308|doi=10.1162/coli_a_00402 |s2cid=219304854 |url=https://doi.org/10.1162/coli_a_00402|doi-access=free|hdl=2078.1/278798|hdl-access=free}}{{cite journal|last1=Nivre|first1=Joakim|last2=de Marneffe|first2=M.C.|display-authors=etal|title=Universal dependencies v1: A multilingual treebank collection|journal=Proceedings of the 10th International Conference on Language Resources and Evaluation (LREC).|date=2016|pages=1659–1666|url=http://www.lrec-conf.org/proceedings/lrec2016/pdf/348_Paper.pdf}} This supports natural language processing, and is a major topic, with its own event, thousands of linguistics and AI researchers working with and on it, and widely-adopted.{{cite web|url=http://universaldependencies.org/|website=Universal Dependencies|title=Universal Dependencies|accessdate=13 December 2016}} The law was put forward by Christopher D. Manning.
The Six Directives
Manning's Law has been described as consisting of six directives,{{cite web|url=http://universaldependencies.org/eacl17tutorial/applications.pdf|website=Universal Dependencies|title=Tutorial on Universal Dependencies at EACL 2017|accessdate=28 August 2018}} which may not necessarily all apply simultaneously, and are often in conflict to some degree:
- UD needs to be satisfactory for analysis of individual languages.
- UD needs to be good for linguistic typology.
- UD must be suitable for rapid, consistent annotation.
- UD must be easily comprehended and used by a non-linguist.
- UD must be suitable for computer parsing with high accuracy.
- UD must provide good support for downstream NLP tasks.
Manning's Law is not the six criteria in themselves, but rather the statement that it is easy to improve UD with respect to a single criterion but hard to improve UD with respect to all criteria at once.