Automatically classifying the role of citations in biomedical articles.

Area of Science:

• Bibliometrics
• Computational Linguistics
• Scientific Communication

Background:

• Citations are fundamental in scientific literature, but traditional models overlook their diverse semantic roles.
• Understanding the specific function of each citation enhances scholarly analysis and knowledge discovery.

Purpose of the Study:

• To develop and evaluate a machine learning model for classifying citation functions within scientific texts.
• To create a robust, eight-category classification scheme for citation roles.

Main Methods:

• Development of an eight-category citation function classification scheme.
• Manual annotation of 1,710 sentences based on the developed schema.
• Training and evaluation of supervised machine learning classifiers on the annotated dataset.

Main Results:

• The developed classifier achieved an average F1-score of 76.5%.
• The study successfully demonstrated the feasibility of automated citation function classification.
• The annotated dataset provides a valuable resource for future research.

Conclusions:

• Automated classification of citation roles is achievable with high accuracy.
• This approach offers a more nuanced understanding of scientific literature compared to traditional citation analysis.
• The classifier is publicly available as a Java API for broader scientific use.