Effective String Processing and Matching for Author Disambiguation

摘要

Track 2 of KDD Cup 2013 aims at determining duplicated authorsin a data set from Microsoft Academic Search. This type ofproblems appears in many large-scale applications that compileinformation from different sources. This paper describes oursolution developed at National Taiwan University to win thefirst prize of the competition. We propose an effective namematching framework and realize two implementations. An importantstrategy in our approach is to consider Chinese and non-Chinesenames separately because of their different naming conventions.Post-processing including merging results of two predictionsfurther boosts the performance. Our approach achieves F1-score0.99202 on the private leader board, while 0.99195 on the publicleader board. color="gray">