I think the idea is, actually, valid. But I don't think the measurement is. I think this plagues all creativity measures: you're asking the subject to be very open in their thinking, but the evaluation method must be narrow and objective.
I would agree: first attempt I scored 74 and analyzing their distance matrix it looks like their measure is broken. e.g. hammer and time are relatively close because "hammer time" is a phrase so their proximity frequency is high, not because the words are semantically related. Second attempt was 89 after a trivial change to account for the broken measure. Nice idea for an experiment but the execution is lacking.