Floating or Suggesting Ideas? A Large-Scale Contrastive Analysis of Metaphorical and Literal Verb-Object Constructions

2026-04-09 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, quick

Summary

A large-scale contrastive analysis investigated how metaphorical and literal verb-object (VO) constructions differ in everyday English. Researchers analyzed 297 English VO pairs, such as "float idea" versus "suggest idea," across approximately 2 million corpus sentences. Utilizing five natural language processing tools, the study extracted 2,293 cognitive and linguistic features, encompassing affective, lexical, syntactic, and discourse-level properties. Cross-pair analysis revealed that literal contexts exhibit higher lexical frequency, cohesion, and structural regularity, whereas metaphorical contexts show greater affective load, imageability, lexical diversity, and constructional specificity. However, within-pair analyses demonstrated significant heterogeneity, indicating that most pairs do not follow uniform patterns, suggesting that differences are largely construction-specific rather than universally consistent.

Key takeaway

For computational linguists developing natural language understanding models, you should account for the construction-specific nature of metaphor-literal distinctions. Relying on a single, consistent distributional pattern to differentiate metaphorical from literal usage will likely lead to inaccuracies. Instead, your models need to incorporate diverse linguistic features and handle the observed heterogeneity in verb-object pair contexts to improve semantic interpretation.

Key insights

Metaphorical and literal language differences are largely construction-specific, not universally consistent.

Principles

• Metaphors express abstract concepts via concrete domains.
• Literal contexts show higher frequency and regularity.
• Metaphorical contexts show greater affective load and diversity.

Method

The study analyzed 297 verb-object pairs in ~2M corpus sentences, extracting 2,293 cognitive and linguistic features using five NLP tools for cross-pair and within-pair comparisons.

In practice

• Analyze specific VO constructions for nuanced meaning.
• Consider affective load in metaphorical language use.

Topics

• Metaphorical Language
• Literal Language
• Verb-Object Constructions
• Natural Language Processing
• Corpus Linguistics

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Probing Semantic Alignment, Lexical Invariance, and Syntactic Influence in LLM Metaphor Processing — LLMs struggle with deep metaphor comprehension, often misinterpreting concepts and relying on lexical associations over context or full syntactic understanding.
• A corpus-assisted metaphor analysis in portraying businesswomen: Diachronic changes in Hong Kong English news — Corpus-assisted metaphor analysis reveals diachronic shifts in media portrayals of businesswomen.
• Beyond semantics: the challenges of annotating pragmatic and discourse phenomena — Annotating abstract semantic and pragmatic information is challenging due to its implicit, context-dependent nature.
• Your AI Probably Understands Your Marriage Joke Better Than Your SQL Parser — LLMs excel at interpreting human language's implied, non-literal meanings, unlike literal parsers.
• Semantic Chunking and the Entropy of Natural Language — A new statistical model explains natural language redundancy through hierarchical semantic chunking and predicts variable entropy rates.
• Modelling the Morphology of Verbal Paradigms: A Case Study in the Tokenization of Turkish and Hebrew — Tokenization strategy critically impacts Transformer models' ability to represent complex morphology, especially in non-concatenative languages.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.