Computational Linguistics 3, Exercise 3: choosing a corpus

Why each is a poor choice on its own, from the Corpus linguistics article:
- The Brown Corpus is about one million words of American English from 1961.
  It is old, American, and small, so it cannot show how people talk about
  money today or in Britain.
- The British National Corpus is a 100 million word collection of spoken and
  written texts created in the 1990s. It is much larger, but it is also dated,
  and British.

The article does not say what either contains about money, so any claim about
their contents on that topic would be a guess.

What I would want instead (my suggestion, not from the sources): a recent
corpus, in the variety I care about, that includes both spoken and written
texts and the kinds of text where money is discussed (news, forums, everyday
conversation), with its date, size and sampling described so I can judge what
my counts represent. The Text corpus article lists web-based corpora of
45-425 million words each (COCA, COHA, TIME, the BNC and others), but I have
not checked what any of them contains or how recent they are, so I name them
only as places to look.

Source: Wikipedia, Corpus linguistics; Text corpus.