The correspondence tables of legislative amendments prepared by the National Diet of Japan mainly focus on the relationships between articles. While this approach effectively ensures legal consistency and transparency, it does not meet the needs of researchers and legal practitioners who require a more detailed understanding of legislative changes. Specifically, they need correspondence tables that illustrate the semantic similarities between articles, providing a more nuanced perspective on the amendments. To address the needs of researchers and legal practitioners, developing machine learning methods to generate correspondence tables based on semantic similarity is essential. Such methods also make it possible to identify complex relationships between provisions that might otherwise be overlooked, especially when considering the significant volume of legislative amendments over time. In this study, we explored the application of computational techniques to automatically create semantic similarity-based correspondence tables, focusing on Japanese Commercial Code. Our results include discovering new relationships between old and new articles that were not apparent in the correspondence table prepared by the National Diet, which focused on the succession of the former. These results highlight the value of this approach as a supplementary tool for legal research and practice. They also reveal the necessity of handling paragraph-level text for corresponding new and old laws.