Automatic literature review generation - Conducting a comparison of research papers

This paper proposes an approach for automatic literature review generation from a set of related research papers, focusing on the similarities and differences among them. The method uses sentence-level and concept-level discourse structure to compare sources and identify important information, followed by a hybrid text generation technique to trim the original sentences and fit them in templates which fulfill the functional purposes of a literature review. The method is illustrated with examples to show how to use semantic representations to identify important concepts and compare them across different sources. This is followed by a description of its implementation in an automatic system.