Machine learning for name type classification in library metadata

Abstract

ABSTRACTThis poster reports on the effectiveness of machine learning approaches to classify common names in library metadata records using the Library of Congress Name Authority File. Features extracted from this dataset were used to train and evaluate classification algorithms including decision tree, naïve Bayes, random forest and support vector machine implemented in Weka, an open‐source machine learning platform. The best performing classifiers were also tested on a collection of 30,000 names extracted from the UNT Digital Library This poster presents the feature sets, their testing results and the information gains of extracted features. The study demonstrated that machine learning could effectively classify names as persons or corporations.