Article level classification of publications in sociology

Ondertitel:an experimental assessment of supervised machine learning approaches

The purpose of this experiment is to assess whether and to what extent it is feasible to make use of supervised machine learning to classify social science journal articles into fine-grained disciplinary categories. Classifying scientific articles according to disciplines is most commonly done by making use of a proxy such as Clarivate Analytics' Web of Science journal level Subject Categories. Past research has shown that this approach does not come without limitations. Classifications based on textual data might be more appropriate in this case. In this paper we make such an attempt using titles and abstracts. We test four different supervised machine learning algorithms and assess their accuracy when it comes to granularly classifying sociology publications based on textual information. Our results show that when the Gradient Boosting model is confronted with unseen test data, it achieves an accuracy which is slightly over 80 percent.

ISBN:978-88-3381-118-5

Jaar van publicatie:2019

Trefwoorden:P1 Proceeding

BOF-keylabel:ja

Authors from:Higher Education

Toegankelijkheid:Closed