Annotation guidelines (version 2.0; UNDER CONSTRUCTION)
Used by the PARSEME corpora annotated for multiword expressions


Categories of MWEs

The top level of MWE categories is motivated by a mixture of morphosyntactic and functional criteria, inspired from the classification of syntactic relations in Universal Dependencies, and includes:

  • verbal MWEs (VMWEs), with several subcategories (defined and annotated in versions 1.0 to 1.3 of these guidelines)
  • nominal MWEs (NMWEs), including nominal idioms and nominal MWEs derived from VMWEs
  • adjectival and adverbial MWEs (AMWEs), including adjectival and adverbial idioms, with separate subcategories for those derived from VMWEs
  • functional MWEs (FuncMWEs), including multiword determiners, adpositions, conjunctions and interjections

This classification, covering all syntactic types of MWEs, is new in version 2.0 of the guidelines. Previous versions covered verbal MWEs only. For a summary of changes with respect to edition 1.3, see the what's new file.

In practice, to identify and categorize MWEs during manual annotation, one must start at the unique entry point and follow the decision diagrams specific for the distribution of a MWE candidate:


An error has occured !