Exploiting Relational Structure to Understand Publication Patterns in High-Energy Physics

Exploiting Relational Structure to Understand Publication Patterns in High-Energy Physics

• Amy McGovern, Lisa Friedland, Michael Hay, Brian Gallagher, Andrew Fast, Jennifer Neville, David Jensen

• Knowledge Discovery Laboratory University of Massachusetts Amherst

Knowledge Discovery Process

Data cleaning and extraction

• Extracted abstracts

• Consolidated authors

• Same name assumed
• 13,185 authors to 9,200
• Co-authored with similar names
• Authors of referenced papers with similar names
• Authors with similar email domains and the same username

Data dependencies

• Examples of high correlations:

• Number of downloads in first 60 days and number of citations
• Is paper published and number of citations (binned)

• Examples of high autocorrelation:

• Journal name (through author)
• Topic cluster of paper (through author)
• Author’s total co-authors (through paper)
• Number of downloads in first 60 days (through journal)

Influential Authors

20% of physicists receive 80% of the citations

Influential authors are more connected

Will a paper be accepted by Physics Letters B?

• Papers from 1995-2000

• 68% accuracy, 0.75 AUC

Identifying Research Communities

Example topic clusters

KDD Cup 2003 Paper: kdl.cs.umass.edu/papers/kddcup2003.html Proximity: kdl.cs.umass.edu/proximity/ Email: [e-mail]