Abstract: In foreign periodicals databases, a prevalent problem is to use the same abbreviation for names of several authors. It seriously affects the accuracy of the author search. This paper attempts to, by utilizing rules and algorithms, enable accurate search by author names: it annotates training data for classification algorithm based on rules, so that supervised algorithm can be conducted in unsupervised conditions. The algorithm is suitable for author name disambiguation of the known authors. Compared with regular disambiguation methods, this method, because of the unsupervised algorithm, does not require manual annotation, and thus features higher efficiency and is easier to correspond with entity. The method is proved to result in higher accuracy in practice.

Key words:  , Author name disambiguation, Data annotation, Classification algorithm, Naive Bayes