After the issue with Indonesian (and Portuguese), it's apparent that the terms localization files are not always themselves clean. We have tests to cover the logic of terms parsing, but maybe we should add a test or two to verify that each localization (and each new one going forward) is working and parsed as expected.