Event coreference resolution on figurative event phrases

Possible course project for incoming students to explore

Normal Example:

The Indian navy has <m> captured </m> 23 Somalian pirates .

Figurative Conversion:

The Indian navy's net of justice has <m> ensnared </m> 23 Somalian sea wolves.
  • “net of justice” metaphorically describes the Indian navy’s operations.
  • “sea wolves” metaphorically represents the pirates, drawing a parallel with wolves which are often considered cunning and predatory creatures.

Research Goals

  • Reuse existing ECR datasets to augment with metaphorical version of the events.
  • Create high-quality examples through multiple rounds of adjudications by also identifying standard and metaphorical versions of the events.
  • Produce baseline results using existing methods. The goal here is produce a dataset on which simple heuristics fail completely, and existing methods score low in performance.

Useful References

Gao, G., Choi, E., Choi, Y., & Zettlemoyer, L. (2018). Neural Metaphor Detection in Context. Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, 607–613. https://doi.org/10.18653/v1/D18-1060

Ichien, N., Stamenković, D., & Holyoak, K. J. (2023). Large Language Model Displays Emergent Ability to Interpret Novel Literary Metaphors (arXiv:2308.01497). arXiv. https://doi.org/10.48550/arXiv.2308.01497

Lai, H., & Nissim, M. (2022). Multi-Figurative Language Generation (arXiv:2209.01835). arXiv. https://doi.org/10.48550/arXiv.2209.01835

Martin, J. H. (1992). Computer Understanding of Conventional Metaphoric Language. Cognitive Science, 16(2), 233–270. https://doi.org/10.1207/s15516709cog1602_4 s10462-023-10564-7.pdf. (n.d.).

Stowe, K., Beck, N., & Gurevych, I. (2021). Exploring Metaphoric Paraphrase Generation. Proceedings of the 25th Conference on Computational Natural Language Learning, 323–336. https://doi.org/10.18653/v1/2021.conll-1.26

Turney, P., Neuman, Y., Assaf, D., & Cohen, Y. (2011). Literal and Metaphorical Sense Identification through Concrete and Abstract Context. Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, 680–690. https://aclanthology.org/D11-1063