PARSEME 2.0 Multilingual Shared Task on Identification and Paraphrasing of Multiword Expressions

• Event title: culminating workshop of the PARSEME 2.0 shared task
• Proposal: submitted to SemEval 2026
• Location: TBA
• Dates: TBA
• Data: to be provided by the ongoing PARSEME/UniDive annotation campaign on multiword expressions
• Shared task organizers:
  • Manon Scholivet, Université Paris-Saclay, France
  • Takuya Nakamura, Université Paris-Saclay, France
  • Agata Savary, Université Paris-Saclay, France
  • Eric Bilinski, Université Paris-Saclay, France
  • Carlos Ramisch, Aix-Marseille Université, France

This subtask is an extension of PARSEME shared tasks on automatic identification of verbal MWEs.

• Task: Given a raw text, automatically underline MWEs in it
• Data: PARSEME 2.0 annotated corpora (not necessarily all the texts from release 1.3)
• Language teams willing to participate with PARSEME data
  • Albanian
  • Egyptian (ca. 2700-2000 BC): MWEs from the UD-EUJA treebank.
  • Georgian
  • Greek (Modern)
  • Greek (Ancient)
  • Hebrew
  • Japanese
  • Lithuanian
  • Persian (Farsi)
  • Polish
  • Romanian
  • Serbian
  • Slovene
  • Swedish
  • Ukrainian
• Minimum annotation effort: 2000 annotated MWEs

• Task: Given a sentence with a MWE, rephrase a sentence so that there is no MWEs but the meaning is the same
• Examples:
  • She made up her mind to… ⇒ She finally decided to…
  • He kicked the bucket ==⇒ He died (But not He passed away)
• Data:
  • Selected sentences form PARSEME annotated corpora
  • The same sentences manually paraphrased
  • One to several hundred examples per language
• Language teams willing to participate
  • Albanian
  • French
  • Greek (Modern)
  • Japanese
  • Hebrew
  • Lithuanian
  • Persian (Farsi)
  • Polish
  • Brazilian Portuguese
  • Romanian
  • Serbian
  • Slovene
  • Swedish
  • Ukrainian

• 31 March 2025 - SemEval proposal submission
• 19 May 2025 - SemEval notification
• 15 July 2025: Sample data ready
• 1 September 2025: Training data ready
• 1 December 2025: Evaluation data ready (internal deadline; not for public release)
• 10 January 2026: Evaluation start
• 31 January 2026: Evaluation end (latest date; task organizers may choose an earlier date)
• February 2026: Paper submission
• March 2026: Notification to authors
• April 2026: Camera ready
• Summer 2026: SemEval workshop (co-located with a major NLP conference)