Mapping Similar Provisions Between Japanese and Foreign Laws

Abstract

A system that automatically maps similar provisions between Japanese and foreign laws is useful in comparative legal research. The purpose of this study is to develop a system for automatically mapping similar provisions. It uses the similarity of document vectors generated by the BERT model to find related provisions. We experimentally found that (1) the BERT model effectively maps similar provisions, (2) the BERT model, which was pre-trained in daily English, falls short in performance for Japanese law translated into English, and (3) LEGAL-BERT, a BERT model specific to the legal domain, may not be effective for English translations of Japanese laws. In addition, we performed mapping between Japanese and foreign laws in English and showed that the performance exceeded simple word-based mapping.