Important citation identification by exploiting content and section-wise in-text citation count.

Area of Science:

• Bibliometrics
• Information Science
• Computer Science

Background:

• Citations are crucial for linking research and assessing academic impact.
• Current methods often treat all citations equally, overlooking nuanced importance.
• Existing approaches for identifying important citations (e.g., citation count, context) have limitations.

Purpose of the Study:

• To develop a novel binary citation classification approach.
• To exploit section-wise in-text citation frequencies, similarity scores, and citation counts.
• To introduce machine learning algorithms for assigning weights to logical sections of research papers.

Main Methods:

• Utilized section-wise in-text citation frequencies, similarity score, and overall citation count.
• Developed a machine learning-based approach to assign weights to logical sections.
• Employed Support Vector Machine, Kernel Linear Regression, and Random Forest for classification.
• Tested on two annotated benchmark datasets (465 and 311 citation pairs).

Main Results:

• The proposed approach achieved improved precision (0.84) compared to the state-of-the-art (0.72).
• Section-specific weighting of citations demonstrated effectiveness in classification.
• Demonstrated the value of incorporating section-level features in citation analysis.

Conclusions:

• The novel approach offers a more refined method for citation classification.
• Section-wise analysis and machine learning weighting enhance the accuracy of identifying important citations.
• This contributes to more effective bibliometric analysis and research trend identification.