A Hybrid approach to recommending universal decimal classification notation

Abstract

In this article we present a hybrid approach to recommending the Universal Decimal Classification notation for unclassified documents. By recommending Universal Decimal Classification notation to librarians, we can enable them to semi-automatically determine the notation using already classified documents. The hybrid approach combines the BM25 method and the naive Bayes classifier, where both methods return a list of recommended notations. Both lists are merged into a final recommendation list using a custom merge function. In detail we present the Universal Decimal Classification notation structure, the corpus of documents, the inputs to our methods and the inner workings of our hybrid approach consisting of both methods. We provide the measurement results of the recommendation lists for the corpus from the National Open-Access Infrastructure in the form of precision, recall and F_β metrics.