Latent Human Error information is gathered in companies, but there is a problem that it cannot be used well. The problem behind it is lack of knowledge on trouble analysis and insufficient analysis on text data. Therefore, in this research, we propose an information analysis system on latent human error. In the analysis system, the following three important elements are included. The first point is knowledge of Human Factors which is useful for trouble analysis. The second point is natural language processing technology which processes text information on trouble. The third point is statistics and DeepLearning technology using processed text information. Based on these factors, we aim to build an analysis support system that can be used by people who are not familiar with trouble analysis at the company site.