Currently, when breaking down characters using IDS data, we use the kZVariant property in Unihan to expand variant characters, which enables us to conduct fuzzy searches. However, there are variant relationships that we want to treat as equivalent even though they are not included in kZVariant. For example, a search for 金 + 斉 should match the character U+2E895 𮢕（⿰釒斉）, but 金 is not related to 釒 through kZVariant.

To expand variant characters appropriately for fuzzy searches, we need to prepare a suitable source for treating the variant characters as equivalent, even if they are not covered by kZVariant.