Research map / Label-efficient and robust learning / Recognition and long-tail
Explainable Metaphor Identification Inspired by Conceptual Metaphor Theory
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add concept-mapping multitask learning to the RoBERTa metaphor classifier (ablation) | Metaphor identification model · MOH; TSV-dev; TSV-test | F1: 73.6 -> 75.6; 85.5 -> 90.2; 74.1 -> 86.6 | Helped |
| Add loss balancing between metaphor and concept-learning losses (ablation) | Multitask metaphor identification model · MOH; TSV-dev; TSV-test | F1: 70.8 -> 75.6; 80.5 -> 90.2; 77.5 -> 86.6 | Helped |
| Use dynamic reward weighting instead of fixed β=0.1 (ablation) | Loss-balanced multitask metaphor identification model · MOH; TSV-dev; TSV-test | F1: 73.9 -> 75.6; 88.1 -> 90.2; 79.6 -> 86.6 | Helped |
| Use generated source and target concepts rather than random concepts (ablation) | RoBERTa concept-feature classifier · GUT-test | F1: 82.4 -> 92.5 | Helped |
| Use generated concepts instead of original word-pair inputs | RoBERTa concept-feature classifier · GUT-test | F1: 98.0 -> 92.5 | Hurt |
| Use learned source concepts rather than randomly selected candidate concepts (ablation) | RoBERTa concept-feature classifier · GUT-test | F1: 91.8 -> 92.5 | Helped |
| Use the proposed method rather than random source candidates | Concept mapping generation · MOH; TSV; GUT; all | accuracy: 56.0 -> 62.0; 63.0 -> 67.0; 59.0 -> 62.0; 59.3 -> 63.7 | Helped |
| Use the proposed method rather than random source candidates | Source concept generation · MOH; TSV; GUT; all | accuracy: 62.0 -> 67.0; 66.0 -> 69.0; 74.0 -> 76.0; 67.3 -> 70.7 | Helped |
| Use 1 rather than 3 most frequent source-word associations (ablation) | Metaphor and concept model · MOH | F1: 71.3 vs. 75.6 | Hurt |
| Use 2 rather than 3 most frequent source-word associations (ablation) | Metaphor and concept model · MOH | F1: 72.8 vs. 75.6 | Hurt |
| Use 4 rather than 3 most frequent source-word associations (ablation) | Metaphor and concept model · MOH | F1: 74.9 vs. 75.6 | Hurt |
| Use 5 rather than 3 most frequent source-word associations (ablation) | Metaphor and concept model · MOH | F1: 72.1 vs. 75.6 | Hurt |
Closest papers by meaning
- CAMEL: Capturing Metaphorical Alignment with Context Disentangling for Multimodal Emotion Recognition (AAAI 2024 · similarity 0.83)
- Nonparametric Identification of Latent Concepts (ICML 2025 · similarity 0.82)
- Concept-Guided Interpretability via Neural Chunking (NeurIPS 2025 · similarity 0.82)
- Redefining <Creative> in Dictionary: Towards an Enhanced Semantic Understanding of Creative Generation (CVPR 2025 · similarity 0.81)
- Seeking the Unfamiliar but Memorable: Conceptual Creativity as Meta-Learning (NeurIPS 2026 · similarity 0.81)
- Any-Way Meta Learning (AAAI 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.