Development of a Lexicon for Pain

Area of Science:

• Computational linguistics
• Medical informatics
• Clinical NLP

Background:

• Pain is increasingly recognized for its association with mental health, posing a significant challenge for natural language processing (NLP) due to its varied textual expression.
• Existing research on pain-related NLP is limited, highlighting a need for specialized resources to accurately interpret clinical text.
• The ambiguity of pain descriptions in text necessitates advanced NLP techniques for clinical applications.

Purpose of the Study:

• To develop a comprehensive English lexicon of pain-related terms for use in natural language processing (NLP) applications.
• To address the challenge of understanding and processing diverse pain descriptions found in clinical and public text sources.
• To create a validated resource that can enhance the accuracy of NLP tasks involving pain identification.

Main Methods:

• Conducted an extensive exploration of pain concepts across diverse text sources, including hospital databases, Twitter, and Reddit.
• Constructed a pain lexicon by deriving terms from scientific literature, existing ontologies, and word embedding models.
• Validated the lexicon through clinical review and comparison with established terminologies like MeSH and a 26-term pain sub-ontology.

Main Results:

• Developed a validated English lexicon comprising 382 pain-related terms.
• The lexicon was derived from literature, ontologies, and word embedding models, ensuring broad coverage and relevance.
• Clinical validation and comparison with existing resources confirmed the lexicon's robustness.

Conclusions:

• The newly developed 382-term pain lexicon is a valuable resource for NLP tasks in healthcare.
• This lexicon will facilitate the selection of relevant documents from electronic health record (EHR) databases and improve pain mention classification.
• The lexicon and associated code are publicly available, promoting further research and development in clinical NLP.