Fine-grained classification of social science journal articles using textual data: A comparison of supervised machine learning approaches

Abstract We compare two supervised machine learning algorithms—Multinomial Naïve Bayes and Gradient Boosting—to classify social science articles using textual data. The high level of granularity the classification scheme used possibility that multiple categories are assigned to a document make this task challenging. To collect training data, we query three discipline specific thesauri retrieve corresponding specialties in classification. resulting data set consists 113,909 records covers 245 specialties, aggregated into 31 subdisciplines from disciplines. Experts were consulted validate thesauri-based multilabel is train algorithms different configurations. deploy classifier chaining model, allowing for an arbitrary number be each document. best results obtained with Boosting. approach does not rely on citation It can applied settings where such information available. conclude fine-grained text-based sciences publications at subdisciplinary hard task, humans machines alike. A combination human expertise suggested as way forward improve documents.