An Ensemble-Based Model to Classify Scholarly Articles on Context: A Path to Recommender System

Articles from multiple research areas tend to use similar words as keywords and titles. On the other hand, selecting a publishing source can be problematic for authors in the initial stage of a scholarly study. So, context-based classification is important where the abstracts, keywords, and titles get similar attention. The aim of this study is to create a tool that uses ensemble learning to classify scholarly articles and recommends sources. This study uses 38 classes for the Web of Science dataset and 40 Classes for the Dimension dataset without grouping them. In all experiment setups models using abstracts achieved the best result as giving a more contextual understanding. Based on this ensemble-based approach, a recommender system has been outlined to recommend probable sources for a given article based on Title, Keywords, and Abstract.