The scorecard solution to the author-paper identification challenge

  • Authors:

  • Xing Zhao

  • Affiliations:

  • FICO, San Diego, CA

  • Venue:
  • Proceedings of the 2013 KDD Cup 2013 Workshop
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper describes team mb74's solution to Track 1 of KDD Cup 2013. The challenge is to determine whether an author has written a given paper in the Microsoft Academic Search database. The key part of our solution is the feature generation which is described with details. A FICO scorecard model with Bernoulli likelihood as the objective is trained in our solution. Alternative objectives and training methods are discussed. It is shown that the scorecard method effectively helps the feature generation and addresses the mean average precision measure for this problem.