Concept analysis is widely used for knowledge foundation. We propose that concept analysis is a reciprocal process rather than sequential, through which keywords taken from relevant documents are enriched. We aimed to facilitate this process by computationally finding out relevant sentences to a given concept. We model a dataset of about 300 PubMed abstracts using a recurrent neural network to build a classifier to predict relevance of a sentence to “diagnostic decision making error”.
Learning Objective: Learn challenges in concept analysis as a spiral process and possible computational solutions using a recurrent neural network.
Teruyoshi Hishiki (Presenter)
Takuro Tamura, University of Tsukuba
Eiji Sugihara, University of Tsukuba
Yoshiko Matsunaga, Toho University
Rie Ichikawa, Nihon University
Jimpei Misawa, Nihon University
Yukihiro Maeda, Nihon University
Akiko Shibuya, Toyoma Prefectural University
Yoshiaki Kondo, Nihon University