The Language of Endometriosis
A Taxonomy of Pain Descriptors and a Rule-Based Language System
Abstract
Pain in endometriosis is frequently described through metaphor, evaluation, and embodied imagery. These descriptions are semantically rich yet difficult to translate into structured clinical documentation. This article presents (1) a linguistically grounded taxonomy of endometriosis pain descriptors derived from corpus-based research, and (2) the design of a rule-based language system that operationalises this taxonomy to generate structured patient and clinician summaries. The paper argues that transparent linguistic modelling offers an interpretable alternative to opaque AI-driven approaches in sensitive health contexts.
1. The Translation Gap in Pain Communication
Endometriosis pain is persistent, cyclical, context-dependent, and multidimensional. Patients frequently rely on metaphor to make invisible sensations communicable:
- “Like broken glass”
- “A tightening band”
- “Electric shocks”
- “Like something burning inside”
Clinicians, however, require structured documentation, categorical clarity, and reproducible summaries. Numeric pain scales capture intensity but not mechanism. Free text preserves experience but resists standardisation. This creates a structural translation gap between lived experience and medical record.
2. Research Foundations
The taxonomy presented here derives from corpus-based analysis of endometriosis pain discourse (≈ 241,000 words). Quantitative analysis showed that pain occurred 2,131 times (≈ 8.8 per 1,000 words), over 120 times more frequently than in the British National Corpus.
31% of instances were figurative. These metaphors were not random; they formed recurring semantic patterns structured around mechanisms such as cutting, burning, pressure, invasion, and entrapment.
The taxonomy follows Conceptual Metaphor Theory and discourse-based evaluation analysis.
3. Taxonomy of Pain Descriptors
The taxonomy is curated, finite, and auditable. Each descriptor is mapped to:
- Semantic domain
- Metaphor category (if applicable)
- Normalised clinical heading
- Interpretive entailment
3.1 Sensory–Physical Domains
| Category | Example Expressions | Type | Semantic Entailment |
|---|---|---|---|
| Cutting Tools | knife, blade, broken glass | Sensory | Pain as penetration or sharp intrusion |
| Pressure / Constriction | tight band, crushing, clamped | Sensory | Pain as compression or restriction |
| Heat | burning, on fire | Sensory | Pain as inflammation or internal heat |
| Electric Force | zapping, electric shock | Sensory | Pain as sudden discharge of energy |
| Weight | heavy, dragging | Sensory | Pain as gravitational burden |
3.2 Emotional and Evaluative Domains
| Category | Example Expressions | Type | Semantic Entailment |
|---|---|---|---|
| Entrapment | trapped, locked in | Emotional | Lack of escape or control |
| Predator / Attack | biting, attacking | Emotional | Pain as external threat |
| Transformation | alien inside me, twisted | Emotional | Bodily intrusion or loss of integrity |
3.3 Contextual Occurrence
| Context | Description |
|---|---|
| Menstruation-related | Cyclical menstrual pain |
| Ovulation-related | Mid-cycle pain |
| Intercourse-related | Dyspareunia |
| Bowel-related | Pain during defecation |
| Persistent / Background | Chronic baseline pain |
4. From Taxonomy to System Design
The taxonomy is operationalised in the Explain My Pain system, a rule-based, deterministic application built with a Flask backend and YAML taxonomy.
User selections are mapped to structured outputs across three dimensions:
- Sensation (mechanism and quality)
- Emotion (evaluative stance)
- Context (trigger or cyclical pattern)
No predictive modelling is used. The system prioritises interpretability, traceability, and ethical transparency.
5. Why Rule-Based Modelling?
In sensitive health contexts, black-box AI systems risk hallucination, overgeneralisation, and loss of patient nuance. A structured taxonomy offers:
- Explicit descriptor → category mapping
- Controlled vocabulary
- Auditable semantic logic
- Reproducible outputs
This does not reject AI. It proposes semantic infrastructure beneath it.
6. Conclusion
The Language of Endometriosis taxonomy demonstrates how applied linguistics can function as infrastructure for digital health systems. Rather than replacing human interpretation, structured modelling supports it.
Interpretability can be designed deliberately. Language can be structured without being flattened.