Vietnamese Author Name Disambiguation for Integrating Publications from Heterogeneous Sources

Abstract

AbstractAutomatic integration of bibliographical data from various sources is a really critical task in the field of digital libraries. One of the most important challenges for this process is the author name disambiguation. In this paper, we applied supervised learning approach and proposed a set of features that can be used to assist training classifiers in disambiguating Vietnamese author names. In order to evaluate efficiency of the proposed features set, we did experiments on five supervised learning methods: Random Forest, Support Vector Machine (SVM), k-Nearest Neighbors (kNN), C4.5 (Decision Tree), Bayes. The experiment dataset collected from three online digital libraries such as Microsoft Academic Search, ACM Digital Library, IEEE Digital Library. Our experiments shown that kNN, Random Forest, C4.5 classifier outperform than the others. The average accuracy archived with kNN approximates 94.55%, random forest is 94.23%, C4.5 is 93.98%, SVM is 91.91% and Bayes is lowest with 81.56%. Summary, we archived the highest accuracy 98.39% for author name disambiguation problem with the proposed feature set in our experiments on the Vietnamese authors dataset.KeywordsDigital LibraryData IntegrationBibliographical DataAuthor DisambiguationMachine Learning