Skip to content

Latest commit

 

History

12 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 

Repository files navigation

MetFuse 🔀

Updated dataset is now available on Huggingface at https://huggingface.co/datasets/CincyNLP/MetFuse


What is MetFuse?

Metonymy and metaphor are two fundamental forms of figurative language — yet they have almost always been studied in isolation. MetFuse is the first dataset to bring them together.

We introduce a framework that takes a literal sentence as input and transforms it into three figurative variants:

Variant What changes Example
Literal The audience clapped when the magician entered the stage.
Metonymy Noun is replaced with a contiguous, in-domain referent The auditorium clapped when the magician entered the stage.
Metaphor Verb is mapped to another domain The audience erupted when the magician entered the stage.
Hybrid Both noun and verb are altered The auditorium erupted when the magician entered the stage.

The dataset contains 1,000 human-verified quadruplets — 4,000 sentences in total — spanning both common noun metonymy (CN) and named-entity metonymy (NE).


Dataset Format

Each instance in MetFuse contains the following fields:

Field Description
sentence The original literal sentence
target_noun The noun that undergoes metonymic substitution
target_verb The verb that undergoes metaphoric mapping
met_par The metonymic paraphrase (the replacement noun)
metonymy The metonymic variant of the sentence
metaphor The metaphoric variant of the sentence
hybrid The hybrid variant (metonymic noun + metaphoric verb)
category Type of metonymy: CN (common noun) or NE (named entity)

Example

target_noun : lawyer
target_verb : see
met_par     : law office
category    : CN

literal     : The lawyer in Lincoln saw issues.
metonymy    : The law office in Lincoln saw issues.
metaphor    : The lawyer in Lincoln beheld the behemoth of issues.
hybrid      : The law office in Lincoln beheld the behemoth of issues.

Download

The MetFuse dataset is available in this repository:

metfuse.csv      ← full dataset (1,000 quadruplets / 4,000 sentences)

Citation

If you use MetFuse in your research, please cite:

@inproceedings{ghosh-jiang-2026-metfuse,
    title = "{M}et{F}use: Figurative Fusion between Metonymy and Metaphor",
    author = "Ghosh, Saptarshi  and
      Jiang, Tianyu",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2026.acl-long.700/",
    doi = "10.18653/v1/2026.acl-long.700",
    pages = "15337--15350",
    ISBN = "979-8-89176-390-6"
}

About

Dataset for the ACL 2026 paper: MetFuse: Figurative Fusion Using Metonymy and Metaphor

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors