@bvibber FYI I moved your library and adjusted it a bit since you last took a look. It now lives within our WMFData package, and it accessible via this WMFWikitextUtils file.

I modified the JS-side a bit to allow passing in targetNamespaces and targetAltParams from Swift, so that it's internationalized.

I quickly slapped together this extractCaptionForDisplay extension method for MissingAltTextLink - I'm calling it after the MissingAltTextLinks are determined before entering the alt text input flow. There may be a cleaner way to do this so I'm happy to scrap it entirely. Basically I just need the caption from your library, or indicators of where the caption starts and how long it is. I can clean it up further in Swift if I need to remove templates, etc. before displaying in the UI. And then later I can use your caption start index for inserting the alt text.

I don't think this is needed for the experiment (which we're hoping to wrap up development this week on), but that could change if QA determines my workaround isn't working well enough.