Towards symbiosis in knowledge representation and natural language processing for structuring clinical practice...

Area of Science:

• Medical Informatics
• Computational Linguistics
• Knowledge Representation

Background:

• Clinical practice guidelines (CPGs) require efficient translation from free text to computable formats for integration into clinical information systems.
• Natural language processing (NLP) offers potential for improving the efficiency of this translation process.
• Developing NLP to structure CPGs using current formal knowledge representations (KR) is labor-intensive.

Purpose of the Study:

• To discuss the value and feasibility of symbiosis between text-based knowledge acquisition (KA) and KR for structuring CPGs.
• To compare a manually created ontology with an automatically derived ontology for CPG eligibility criteria.
• To explore interweaving KA and NLP for KR and identify considerations for achieving symbiosis.

Main Methods:

• Comparison of two ontologies: one expert-created for CPG eligibility criteria, and one automatically derived from a semantic pattern-based approach.
• Discussion of the strengths and limitations of combining KA and NLP for KR.
• Exploration of symbiosis between KR and NLP for structuring CPGs.

Main Results:

• The paper presents a vision for symbiosis between KA and KR, leveraging NLP for efficient CPG structuring.
• It highlights the trade-offs between manual and automated ontology development.
• Identifies key considerations for successful integration of KR and NLP.

Conclusions:

• Symbiosis of KA and NLP is valuable and feasible for structuring CPGs into computable formats.
• This approach can enhance the adoption of evidence-based CPGs in clinical practice.
• Further research is needed to optimize the interweaving of KA and NLP for KR.