Predicting the Confidential Words in Japanese Judicial Precedents Using the Neural Network Combined POS Tag

Abstract

In Japanese judicial field, most precedents are not publicly opened on Japanese court web pages to protect personal information. Confidential words such as personal names are replaced by other meaningless words. This operation takes time and effort because it is done manually. Therefore, we would like to predict the confidential words automatically. We propose the method of predicting the confidential words in Japanese judicial precedents using the neural network combined POS (Part of speech) tag. As the result we obtained 88% accuracy improvement for detecting the confidential words compared to the previously model. Then, we confirmed whether our model could be practical use or not.