Processing of figurative language is a rapidly growing area in NLP, including computational modeling of metaphors, idioms, puns, irony, sarcasm, simile, and other figures. Characteristic to all areas of human activity (from poetic, ordinary, scientific, social media) and, thus, to all types of discourse, figurative language becomes an important problem for NLP systems. Its ubiquity in language has been established in a number of corpus studies and the role it plays in human reasoning has been confirmed in psychological experiments. This makes figurative language an important research area for computational and cognitive linguistics, and its automatic identification, interpretation and generation indispensable for any semantics-oriented NLP application.
The proposed workshop will be the fourth edition of the biennial Workshop on Figurative Language Processing, whose first editions were held at NAACL 2018, ACL 2020 and EMNLP 2022, respectively. The workshop builds upon a long series of related workshops that the current organizers have been involved with: “Metaphor in NLP” series (2013-2016) and “Computational Approaches to Linguistic Creativity” series (2009-2010). We expand the scope to incorporate various types of figurative language, with the aim of maintaining and nourishing a community of NLP researchers interested in this topic. The main focus will be on computational modeling of figurative language, however papers on cognitive, linguistic, social, rhetorical, and applied aspects are also of interest, provided that they are presented within a computational, formal, or a quantitative framework. Recent advancement in language models have led to several works on figurative language understanding (Chakrabarty et al 2022a; Chakrabarty et al 2022b; Liu et al 2022; Hu et al 2023) and generation (Stowe et al 2021; Chakrabarty et al 2021; Sun et al 2022; Tian et al 2021) At the same time large language models have opened up opportunities to utilize figurative language in scientific (Kim et al 2023) as well as creative writing (Chakrabarty et al 2022c; Tian et al 2022). Additionally there have also been recent work on multimodal figurative language generation (Chakrabarty et al 2023; Akula et al 2023), understanding (Hessel et al 2023; Yosef et al 2023) and interpretation (Hwang et al 2023; Desai et al 2022; Kumar et al 2022). We encourage submissions along these axes.
Topics of Interest
The workshop will solicit both full papers and short papers for either oral or poster presentation. Topics will include, but will not be limited to, the following:
Identification and interpretation of different types of figurative language: Linguistic, conceptual and extended metaphor; irony, sarcasm, puns, simile, metonymy, personification, synecdoche, hyperbole
Generation of different types of figurative language: sarcasm, simile, metaphors, humor, hyperbole
Multilingual and multimodal figurative language processing
Resources and evaluation
Annotation of figurative language in corpora
Datasets for evaluation of tools
Evaluation methodologies
Figurative use in low-resource languages
Processing of figurative language for NLP applications
Figurative language in sentiment analysis; dialogue systems; computational social science; educational applications
Figurative language and mental health
Figurative language in digital humanities
Figurative language in creative writing
Figurative language and cognition
Cognitive models of processing of figurative language by the human brain
Human-AI collaboration for figurative language
Shared Tasks
Multilingual euphemisms detection: Euphemisms are a linguistic device used to soften or neutralize language that may otherwise be harsh or awkward to state directly (e.g. "between jobs" instead of "unemployed", "late" instead of "dead", "collateral damage" instead of "war-related civilian deaths"). By acting as alternative words or phrases, euphemisms are used in everyday language to maintain politeness, mitigate discomfort, or conceal the truth. While they are culturally-dependent, the need to discuss sensitive topics in a non-offensive way is universal, suggesting similarities in the way euphemisms are used across languages and cultures. We propose a shared task in which participants will need to disambiguate sentences in multiple languages as either euphemistic or not. The dataset will include English, Mandarin, Spanish, Yoruba, and possibly additional languages.
Understanding of Figurative Language through Visual Entailment: One important modality that has gained interest recently is vision, namely the interpretation of figurative language in media such as memes, art, or comics. This task is challenging because it involves reasoning abstractly about images, and also involves understanding social commonsense and cultural context. We will frame this as a visual entailment task where a model not only has to predict if a caption entails the content in the image but also provide free text explanations justifying the label prediction. These tasks have proved difficult for state-of-the-art multimodal models in the past. We will have a paper and a baseline for the same.
Important Dates
Long, Short & Demonstration Paper Submission: March 10th, 2024
Long, Short & Demonstration Paper Notification: April 14th, 2024
Final Paper Submission: April 24th, 2024
Workshop: June 21/22, 2024