Research map / Vision-language and multimodal / Human and embodied modeling
Modelling Form-Meaning Systematicity with Linguistic and Visual Features
Arie Soeteman, Dario Gutierrez, Elia Bruni, Ekaterina Shutova
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| optimize edit weights (SMLKR) instead of unweighted Levenshtein (ablation) | text-based model · form-meaning correlation | Mantel correlation: 0.0362 -> 0.0383 | Improved |
| optimize edit weights instead of unweighted (ablation) | image-based model · form-meaning correlation | Mantel correlation: 0.0198 -> 0.0243 | Improved |
| optimize edit weights instead of unweighted (ablation) | scoring-level fusion · form-meaning correlation | Mantel correlation: 0.0401 -> 0.0420 | Improved |
| optimize edit weights instead of unweighted (ablation) | neural network fusion · form-meaning correlation | Mantel correlation: 0.0175 -> 0.0266 | Improved |
| add visual features via scoring-level fusion | text-based model · form-meaning correlation | Mantel correlation: 0.0383 -> 0.0420 | Improved |
| add visual features via feature concatenation | text-based model · form-meaning correlation | Mantel correlation: 0.0383 -> 0.0376 | No clear effect |
| add visual features via Siamese neural network fusion | text-based model · form-meaning correlation | Mantel correlation: 0.0383 -> 0.0266 | Hurt |
| use image-based vectors instead of text-based | form-meaning model · form-meaning correlation | Mantel correlation: 0.0383 -> 0.0243 | Hurt |
| use image-based instead of text-based vectors | phonaestheme detection · phonaesthemes p < 0.05 | count: 22 -> 15 | Hurt |
| use multimodal concatenation instead of text-based | phonaestheme detection · phonaesthemes p < 0.05 | count: 22 -> 10 | Hurt |
| use neural network fusion instead of text-based | phonaestheme detection · phonaesthemes p < 0.05 | count: 22 -> 7 | Hurt |
| combine text and image distances with alpha (ablation) | scoring-level fusion · form-meaning correlation | Mantel correlation: 0.75 | Improved |
Closest papers by meaning
- Textural or Textual: How Vision-Language Models Read Text in Images (ICML 2025 · similarity 0.85)
- Do Language Models Associate Sound with Meaning? A Multimodal Study of Sound Symbolism (AAAI 2026 · similarity 0.85)
- Unveiling Visual Perception in Language Models: An Attention Head Analysis Approach (CVPR 2025 · similarity 0.84)
- ASCIIEval: Benchmarking Models' Visual Perception in Text Strings via ASCII Art (ICLR 2026 · similarity 0.83)
- Invariant Features in Language Models: Geometric Characterization and Model Attribution (NeurIPS 2026 · similarity 0.83)
- Evaluating Semantic Variation in Text-to-Image Synthesis: A Causal Perspective (ICLR 2025 · similarity 0.82)
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.
