Figurative language such as metaphors, similes, sarcasm, or humor is often conveyed visually, and frequently appear in advertising, news, and social media. In the previous iteration of the workshop, we introduced a shared task for figurative language understanding around a textual entailment paradigm, where the hypothesis is a sentence containing the figurative language expression (e.g., metaphor, sarcasm, idiom, simile) and the premise is a literal sentence containing the literal meaning. In this shared task, we aim at Visual Understanding of Figurative Language framed as a visual entailment task: given an pair, a model needs to predict whether the image Entails or Contradicts the claim. The figurative component can be in the textual or in the visual component of the entailment pair.
This task contains a compilation of datasets including visual metaphors, idioms, similes, sarcasm and humor. See details in the Data section. There are two important aspects of this task and the associated dataset: 1) the task requires not only to generate the label (entail/contradict) but also to generate a plausible explanation for the prediction; 2) the entail/contradict label and the explanation are related to the meaning of the figurative expression.
This shared task will be part of the 4th Workshop on Figurative Language Processing. The workshop will be held in conjunction with NAACL 2024, and will be a hybrid (in-person and virtual) event.
Important dates:
02/02/24: Competition opens; participants develop approaches on training data and labels
02/25/24: Test data is released; performances will be tracked on Codabench dashboard
03/03/24: Last day to submit predictions on test data
03/10/24: Papers describing approaches are due
04/14/24: Notification of acceptance
04/24/24: Final paper submission
06/20/24 or 06/21/24: Workshop