Understanding Similes in Text: A Survey of Machine Learning and Deep Learning Approaches for Figurative Language Recognition

Abstract

Figurative language is the bedrock of human com- munication, and texts are enriched with expressions that go beyond literal meanings. Among its diverse forms, similes stand out for their ability to draw vivid comparisons, enhancing comprehension and creativity. This survey paper explores the intersection of simile detection and interpretation with ad- vancements in machine learning (ML) and deep learning (DL). Machine learning, an artificial intelligence subcategory, is the use of statistical techniques to enable computers to learn from data, whereas deep learning, a special type of ML, utilizes neural networks to achieve outstanding accuracy in complex tasks. This paper reviews the state-of-the-art ML and DL methods used for simile analysis in figurative language with respect to their architectures, datasets, and evaluation metrics. We focus on prob- lems of figurative construct processing, which include ambiguity and context dependency, and argue how current computational techniques overcome the problems mentioned above. Drawing on linguistic and AI-based approaches, this paper endeavors to explain how technological breakthroughs in similes relate to how figurative language is now being studied with a difference.