Open research knowledge graph for structuring scholarly contributions using transformers

More research papers are being published now than have ever been at any point in history. It is becoming increasingly difficult for the researchers to keep up with the papers that are being published in even a very narrow domain. This study proposes to build an open research knowledge graph (ORKG) that shows the scholarly contributions of the published papers. The paper makes use of natural language processing techniques and state-of-the-art deep learning models to achieve this task. The system generates a knowledge graph after performing four main steps including sentence classification, phrase extraction, triple formation (and classification) and finally, knowledge graph generation. Different state-of-the-art deep learning models such as RoBERTa have been used for classification and phrase extraction tasks whereas triple formation was performed using different heuristics. Finally, a knowledge graph is generated through which an end-user can identify the scholarly contributions in scholarly article. Experimental results are compared against other systems and show encouraging results.