A good critique by Russ Wilson of the Futures Company of the limitations of tag clouds to actually explain things:
I have two main issues with Wordles [a tag cloud generator], and they’re exemplified in the wordle above, based on David Cameron’s coalition speech. First, they remove the word from its immediate context. Take the word interest, represented as one of the more frequently occurring words. But it could equally indicate curiosity and engagement or interest payments.
The second issue is that frequency is being proposed as an indicator of importance, but that’s not how we actually interpret speech.... Frequency of use is simply that – frequency of use.
Both of these are classic problems in information management-- or to use the old school term, indexing. When I was at Britannica, people often used the example of "depression" in the same way Wilson talks about interest: the Britannica had articles on the geological kind, the economic phenomenon, and the psychological condition, but it was hard for most automated systems to distinguish between them. Users, in contrast, could do so quickly-- because they brought (or could quickly surmise) the specific context in which an instance of the word appeared.
And a moment's reflection should make clear that frequency and importance are not the same thing. Sometimes they are. But how many times is the word "like" over-used to the point of meaninglessness-- ether as a verbal tic, or, thanks to the flattening of the term by Facebook, as the Web equivalent of Valley girl-speak?