Annotating Pain Language: Integrating Linguistics and NLP

CMT + MIP → an NLP-ready annotation schema with clinical impact

Dr Stella Bullo

1. Introduction

Pain is invisible and subjective, which makes it hard to describe. Literal phrases like “sharp pain in my lower back” are clinically useful but often miss the lived quality of the experience. Patients therefore reach for metaphors—e.g., “barbed wire around my spine”, “a hammer pounding inside my head”—to translate sensations into shared domains (objects, impacts, heat).

For linguistics and NLP, the challenge is twofold: detection of metaphorical use with reliability and representation in structured, model-ready data. This article combines Conceptual Metaphor Theory (CMT) and the Metaphor Identification Procedure (MIP) into a compact, actionable annotation schema.

2. Method Foundations

2.1 Conceptual Metaphor Theory

CMT explains how abstract experiences are framed via concrete domains. In pain discourse we see recurring frames:

  • Pain as object“barbed wire around my spine”
  • Pain as impact“a hammer pounding inside my head”
  • Pain as heat“a knot of fire in my stomach”

These categories provide a taxonomy for annotating metaphors.

2.2 Metaphor Identification Procedure

  1. Read the text to establish overall meaning.
  2. Identify lexical units.
  3. Determine each unit’s basic meaning that is more concrete or older.
  4. Compare with its contextual meaning in the utterance.
  5. If they contrast but are understood by comparison, mark as metaphorical.

2.3 Worked example

“It feels like barbed wire wrapped around my spine.”
  • Units barbed wire, wrapped, spine
  • Basic fencing material with barbs; physically enclosed; bones of the back
  • Contextual sharp, constricting pain; enclosing sensation; literal anatomical reference
  • Decision barbed wire and wrapped → metaphorical; spine → literal. Category pain as object

3. Annotation schema

Minimal fields that balance linguistic richness and computational tractability

  • type metaphor or literal
  • trigger word or phrase signalling pain
  • category conceptual frame if type is metaphor
  • location body part

4. Example annotations

TextTypeTriggerCategoryLocation
It feels like barbed wire wrapped around my spine.metaphorbarbed wirepain as objectspine
There is a hammer pounding inside my head.metaphorhammer poundingpain as impacthead
My stomach is a knot of fire.metaphorknot of firepain as heatstomach
I have a sharp pain in my lower back.literalsharp painlower back
I feel pressure in my chest when I breathe.literalpressurechest
{
  "text": "It feels like barbed wire wrapped around my spine.",
  "type": "metaphor",
  "trigger": "barbed wire",
  "category": "pain as object",
  "location": "spine"
}

5. Applications

  1. Automatic metaphor detection annotated corpora to train models that identify figurative language in notes and patient forums
  2. Clinical communication tools mapping metaphors to standardised terminology such as hammer pounding to severe pulsating headache
  3. Explainable healthcare AI metaphor categories as interpretable features in decision support
  4. Cross linguistic research comparative studies informing culturally sensitive multilingual systems
  5. Training and education annotated cases for clinician education and empathy building

6. Bibliography

  • Lakoff, G., and Johnson, M. 1980. Metaphors We Live By. University of Chicago Press.
  • Pragglejaz Group. 2007. MIP A Method for Identifying Metaphorically Used Words in Discourse. Metaphor and Symbol 22 1 1–39.
  • Semino, E. 2010. Describing pain the role of metaphor. Metaphor and Symbol 25 4 250–269.
  • Steen, G. 2017. Deliberate metaphor theory. John Benjamins.
  • Wallington, A. et al. 2003. Metaphor annotation a case study. In Proceedings of the Workshop on Computational Lexical Semantics ACL.

7. Conclusion

CMT plus MIP provide a rigorous and reproducible basis for annotating pain metaphors. The resulting structured data supports better models and clearer clinical tools at the point where linguistics and NLP meet practice.