In the last batch there were a few words that I was vaguely confident of but a lot more of them seemed like "stunt" words I would never see because every time they'd need defining so why bother.
Also I was assuming it was picking from a huge set, but it seems everybody was shown the same words, so while it's supposedly a "sample" any bias, even if unintended, shows up in the results, if you wanted to be scientific perhaps you'd do this for 1000 words and then sample 100 questions from that for each participant or something.