Discovering IMRaD Structure with Different Classifiers

Information within published papers around the world in scientific journals are structured in the format of Introduction, Methodology, Results, and Conclusion (IMRaD). Human ability to read and analyze is not capable of processing these large amounts of information. If we could identify the structure and consequently extract it to a user who needs a part of the structure, particularly an article in a foreign language, time will be saved as result. Computational approaches like Machine Learning (ML) and Natural Language Processing (NLP) have been widely used for similar purposes. However, it is very important to identify which one, or which group of classifiers work better for a specific kind of problem. The objective of this work is to identify applicable classifiers by analyzing and comparing results produced by different ML classifiers used in locating and classifying sentences from abstract of a paper into the IMRaD structure. This work demonstrates the possibility of integrating ML and NLP for the articles' sentence classification based on the IMRaD structure. It also verifies that it is possible to achieve good results with simple implementations without the need of too many computational resources.