The source they used (freeDictionary) does all of the work of parsing wiktionary and giving you simple json object for each word. (It actually started life as a way to get the definitions out of Google's "define:" operator, but it seems those days are over)
It also, turns requests for some words, like "rolling" into definitions for their root word like "roll", even when wiktionary has distinct and useful definitions for the word, which makes it less than ideal for me.
I wondered the same. The main downside is that you need to do some processing to extract the entries from the dump and get the plain text of the fields you want.
I'm also a little surprised they didn't think Wiktionary was sufficient for languages apart from English. I could be wrong, but my impression is that it's pretty good for major languages[1].
