Expertise in French health forums

Area of Science:

• Computational linguistics
• Medical informatics
• Natural language processing

Background:

• Health websites increasingly feature medical experts to ensure answer quality.
• Medical professionals often contribute to online health forums without explicit role identification.
• Identifying expert-authored content is crucial for accessing accurate health information.

Purpose of the Study:

• To develop and evaluate machine learning models for detecting posts authored by medical experts in health forums.
• To identify optimal features and methods for text categorization in this domain.

Main Methods:

• Utilized text classification techniques on data from two French health forums.
• Employed a Random Forest classifier to learn patterns distinguishing expert from non-expert posts.
• Analyzed misclassified posts to understand limitations and nuances.

Main Results:

• Achieved high performance, with over 98% F1-measure using the Random Forest classifier.
• Demonstrated that models trained on one forum can generalize effectively to others.
• Highlighted that medical experts participate actively even when their roles are not explicitly stated.

Conclusions:

• Automated detection of medical expert contributions in health forums is feasible and effective.
• This technology can significantly improve the reliability and accessibility of online health information.
• Further research can refine models to better capture implicit expert identification.