Mining consumer health vocabulary from community-generated text

Area of Science:

• Natural Language Processing
• Medical Informatics
• Computational Linguistics

Background:

• Community-generated text corpora, like Wikipedia, are rich sources for understanding consumer health vocabulary (CHV).
• Linking CHV to professional medical terminology is crucial for improving health information systems.
• Existing methods may not effectively differentiate between consumer and professional health terms in large datasets.

Purpose of the Study:

• To propose and evaluate a pattern-based text-mining approach for identifying and classifying consumer health vocabulary (CHV) from community-generated text.
• To develop a novel frequency-based measure for distinguishing consumer terms from professional medical terms.
• To assess the accuracy of the proposed approach in identifying synonymous term pairs.

Main Methods:

• Utilized Wikipedia as a large, community-generated text corpus.
• Developed a pattern-based text-mining strategy to extract potential CHV and professional term pairs.
• Implemented a novel frequency of occurrence ratio to differentiate between consumer and professional terminology.
• Empirically validated the approach using MedLine abstracts and online health forum data (MedHelp).

Main Results:

• The proposed approach successfully identified synonymous pairs of consumer and professional health terms.
• The method demonstrated high accuracy in labeling terms as either consumer or professional.
• The frequency-based measure effectively differentiated between the two types of health vocabulary.
• The approach showed applicability across diverse data sources, including scientific abstracts and forum posts.

Conclusions:

• The pattern-based text-mining approach is effective for extracting high-quality consumer health vocabulary (CHV).
• This method has significant potential to enhance computational applications dealing with consumer-generated health text.
• Accurate CHV extraction improves the performance of systems processing patient-generated health information.