Well, although this trick works, it would be a much better idea to regenerate the data files to be used with the newest Mbt. (main improvement being that it better handles Unicode issues.
I did find the data-file for dum, but cannot access it (@proycon could you change the right on the server @ru ?)
I didn't yet find the original data for nld-vnn. I'm not sure where it stems from. @proycon any idea?

@tcbrouwer For the time being, your hack is sufficient. A regenerated Mbt file most likely just gives minor different results (like in Confidence values)

@tcbrouwer Thanks for the pull request. I didn't realize that this was now broken out-of-the-box, we definitely need to release a fixed frogdata then. Your fix looks like a quick and acceptable one, unless we indeed want to regenerate it properly like @kosloot suggests.

(@proycon could you change the right on the server @ru ?)

Done!

The best source however is https://github.com/INL/nederlab-linguistic-enrichment/tree/master/resources (it's a private repo at INT but you probably still have access there from back then). This was all done in the scope of the Nederlab project in 2018.

I didn't yet find the original data for nld-vnn. I'm not sure where it stems from. @proycon any idea?

That's this one frog-bab-cgn.

I think the source materialis non-free unfortunately, hence the private repo.

the 'dum' data is now updated.
nld-vnn is still to be done

This is now releasedin frogdata v0.21 , but nld-vnn remains to be done so the current state is not ideal