Extracting Characteristics of the Study Subjects from Full-Text Articles

Area of Science:

• Biomedical Informatics
• Information Retrieval
• Medical Indexing

Background:

• Biomedical research subject characteristics are crucial for publication relevance.
• MEDLINE citations use Medical Subject Headings (MeSH) for subject description.
• Current Medical Text Indexer (MTI) recommendations for MeSH can be improved.

Purpose of the Study:

• To enhance the Medical Text Indexer's (MTI) recommendation of Medical Subject Headings (MeSH).
• To explore the utility of full-text content for improving MeSH indexing.
• To assess the impact of incorporating methods sections and captions on indexing accuracy.

Main Methods:

• Extracted sentences from methods sections and captions of research publications.
• Augmented existing abstracts with extracted full-text sentences.
• Processed augmented abstracts using the Medical Text Indexer (MTI).
• Evaluated indexing performance using recall, precision, and F1 score.

Main Results:

• Augmenting abstracts with full-text sentences significantly improved recall and F1 score.
• A slight decrease in precision was observed with the augmented abstract approach.
• Directly assigning headings from full text also yielded improvements.
• The study demonstrated the value of leveraging full-text data for indexing.

Conclusions:

• Incorporating full-text data, specifically from methods and captions, enhances MeSH heading recommendations.
• Automated methods for utilizing full-text content are essential for advanced indexing.
• Further development is needed to optimize the use of full-text information in biomedical literature indexing.