A deep learning approach for identifying biomedical breakthrough discoveries using context analysis

Abstract Breakthrough research in scientific fields usually comes as a manifestation of major development and advancement. These advances build to an epiphany where new ways thinking about problem become possible. Identifying breakthrough can be useful for cultivating funding further innovation. This article presents method identifying breakthroughs from papers based on cue words commonly associated with advancements. We looked specific terms signifying citing sentences identify articles. By setting threshold the number (“citances”) that peer scholars often use when evaluating research, we identified articles containing research. call this approach “others-evaluation” process. then shortlisted candidates selected authors’ evaluations their own found abstracts. “self-evaluation” Combining two approaches into dual “others-self” evaluation process, arrived at sample 237 potential articles, most which are recommended by Faculty Opinions. Based identified, using SVM, TextCNN, BERT train models abstracts evaluations. automatic identification model greatly simplify process others-self-evaluation promote