Metaphors shape how we describe ambition, responsibility, guilt, illness, and political change. They are cognitively powerful but computationally difficult. A robust modelling approach requires more than keyword matching; it requires structured linguistic design.
Key idea
Effective metaphor analysis requires infrastructure: human annotation → structured taxonomy → rule-based tagging → interpretable output.
1 · Schema-Driven Annotation
Most annotation platforms prioritise speed and label collection. Linguistic modelling requires something different: explicit theoretical categories that preserve interpretive assumptions.
- Fields derived from metaphor theory and stance analysis
- Encoding of
phenomenon,stance_tone, andtranslation_shift - Version-controlled schemas preserving modelling logic
Categories are not neutral. They shape what researchers can observe and what systems can learn. A schema-driven interface makes those assumptions visible.
2 · Rule-Based Metaphor Tagger
The Metaphor Tagger formalises conceptual domains such as FIRE (intensity), WEIGHT (burden), and SHARP OBJECTS (intrusion). Detection rules distinguish figurative from literal usage through contextual constraints.
- Taxonomy derived from corpus-based metaphor research
- Rule logic informed by narrative analysis
- Evaluation using Precision, Recall, F1, and agreement metrics
Annotation App
↓
Gold-standard dataset
↓
Rule-based tagger
↓
Evaluation (Precision / Recall / F1 / κ)3 · Human-Facing Application
Explain My Pain demonstrates how modelling infrastructure becomes interpretable output. Model results are translated into accessible explanations for clinicians and patients.
- Highlight metaphor usage
- Explain semantic entailments
- Support clearer communication
Localisation Layer
Extending the pipeline across English and Spanish enables comparative metaphor modelling. Parallel annotation allows analysis of translation shifts, register variation, and conceptual divergence.
EN / ES parallel corpus
↓
Schema with translation_shift field
↓
Tagger (EN + ES)
↓
Cross-cultural comparisonConclusion
Metaphor modelling is not a single tool but a structured workflow. When annotation design, taxonomy construction, and rule logic are aligned, linguistic theory becomes executable infrastructure.
This pipeline demonstrates how qualitative insight can be formalised, evaluated, and translated into systems that remain transparent and interpretable.