[1N01-07] PRA Method

[1N07] A Study of Natural Language Processing for Classification of Japanese Reports

*Kento Tsukamoto¹, Takanori Kitada¹, Satoshi Takeda¹ (1. Osaka Univ.)

Keywords:Classification of reports, Natural language processing, Machine learning, Probabilistic risk assessment

For probabilistic risk assessment, methods have been proposed to quantitatively evaluate human involvement and interdependence of components by analyzing many reports. On the other hand, Natural Language Processing (NLP) technology using deep learning has the potential to streamline the work and research of analyzing reports written in natural language. The purpose of this research is to support the task of labeling reports written in Japanese. We investigated a method to construct an NLP model for that purpose.