Data Taxonomy for Pain Language

~5–6 min · Updated Sep 2025

Introduction

Pain is often described in messy, unpredictable ways. A patient might say “it burns here,” “it feels like a hammer,” or “my back screams at night.” These expressions are clinically meaningful, yet without a structure they are difficult to compare, tag, or process.

This article presents a compact data taxonomy for pain language, built from my academic linguistic research in the Language of Endometriosis Project. Working directly with patient narratives revealed recurring patterns that annotators can label consistently and developers can map into tools like the prototype pain tagger and the Explain My Pain app.

What is a taxonomy

In annotation, a taxonomy is a “menu” of categories. Instead of asking “what do I feel this means?”, annotators choose from clearly defined options. This reduces confusion, improves agreement, and enables downstream computation.

Why keep it compact

Oversized schemas slow teams and reduce consistency. For pain language, I deliberately kept the taxonomy small and practical—broad enough to capture variety, specific enough to be recognized quickly.

The categories

Illustrated with examples from the Language of Endometriosis corpus.

1) Pain qualities

What the pain feels like: burning, stabbing, throbbing, dull.
Example: “It feels like a burning rope around my stomach.”

2) Body location

Where pain occurs: womb, lower back, knee, stomach.
Example: “A sharp pain in my lower back.”

3) Intensity markers

How strong it is: mild, severe, unbearable, very.
Example: “It’s unbearable every morning.”

4) Figurative or metaphorical language

Comparisons to other domains: knife, hammer, beast, fire.
Example: “It’s as if someone is stabbing me with a knife.”

5) Temporal markers

When it happens or changes: constant, comes and goes, at night, before period.
Example: “The cramps come and go, but always return before my period.”

Entailments in metaphorical pain language

Beyond tagging metaphors, we can capture their entailments—the logical consequences a metaphor implies. In the Language of Endometriosis data, metaphors structured how patients understood and lived their pain.

Metaphor typeExampleEntailments
Violence / Attack “Like being stabbed with a knife.” External agent; patient as victim; aggression; relentlessness.
Weight / Pressure “A heavy rock pressing on my pelvis.” Constancy; immobilization; oppression; limited movement.
Animal “A beast clawing at my womb.” Unpredictability; lack of control; hostility; ‘alive’ pain.
Heat / Fire “Flames burning through my abdomen.” Escalation; spread; destruction; consuming quality.
Containment / Trapping “Being locked in a burning cage.” Restriction; enclosure; no escape; surrounding pain.

Entailments add conceptual depth for analysis and can be mapped to higher-level “pain profiles” in the pain tagger and the app.

Relations

In the Endometriosis data, metaphors often co-occurred with location and intensity; capturing relations keeps the representation faithful to patient speech.

Why this matters

For annotators: a trainable, testable menu. For recruiters: lean design scales better and lowers costs. For developers: a bridge from narrative to features that power the tagger and the app.

Closing

This compact taxonomy is a foundation. Next in the series: Lightweight QA & Sampling, the Metaphor to Measurement case study, and a toolkit of Annotation Snippets & Prompts.