Public repository with code & data for the PARSEME shared task.


Welcome to the data repository of the PARSEME Shared Task on Verbal Multi-Word Expressions Identification.

This repository contains corpora in multiple languages. Corpora are human-annotated with occurrences of Verbal Multiword Expressions (VMWEs).

You can download this repository as a ZIP file.


Human annotations can be found for multiple language directories, such as FR for French, DE for German, PL for Polish, etc. Inside each language directory, you may find these files:

  • A description of the available data for that given language.
  • trial.parsemetsv: Trial file with an example of the final data (released mid-December/2016).1
  • train.parsemetsv: Training data (to be released early January/2017).
  • test.blind.parsemetsv: The blind test data (to be released by the end of January/2017).
  • test.parsemetsv: The gold test data (to be released after the shared task)
  • Number of sentences, tokens and annotated VMWEs in train/test.

The PARSEME-TSV files contain human annotation of VMWEs. Depending on the language, these files may also be paired with a CoNLL-U file for extra information (check the language-specific files).

Note: For some languages, the UPOSTAG and DEPREL fields of their CoNLL-U file may contain data that is not in the universal UD format. We thought it would be useful to provide this information nevertheless.


The bin directory contains useful scripts.

  • bin/ script for evaluating machine predictions against human annotations.
  • bin/ script for checking if the prediction file is in the proper format.

  1. Trial data was removed from the current release in order to avoid confusion.