It would not be too hard to add a field for an IPA version of the name. But populating the dataset would be a huge chicken-and-egg problem.