Semantic Space models for classification of consumer webpages on metadata attributes

Area of Science:

• Computational linguistics
• Health informatics
• Machine learning

Background:

• Online health information presents challenges in quantity and quality.
• Web portals can curate reliable resources for specific health topics or user communities.

Purpose of the Study:

• To develop and evaluate automated methods for classifying online health webpages.
• To assess the accuracy of machine learning models in categorizing content based on language patterns.

Main Methods:

• Utilized Hyperspace Analogue to Language (HAL) to create Semantic Space models of webpage language.
• Applied machine learning algorithms: Support Vector Machine (SVM), Decision Forest, and Summed Similarity Measure (SSM).
• Classified webpages based on metadata attributes within the Breast Cancer Knowledge Online portal.

Main Results:

• Achieved over 93% accuracy in classifying 'medical' vs. 'supportive' perspectives.
• Reached over 92% accuracy for 'early' vs. 'advanced' disease stages.
• Attained over 90% accuracy in distinguishing 'lay' vs. 'clinician' author credentials.

Conclusions:

• Language use patterns in webpages can be effectively modeled using Semantic Spaces.
• Automated classification of online health content is feasible with high accuracy.
• This approach can enhance the organization and accessibility of reliable online health information.