An up to date keyword set on global news

[acszs]

My question is not strictly binds to the topic of text mining, but maybe you can help.
I am hunting for a keyword set, which has the following criterions:
- contains only english words/n-gramms or named entities
- manual (tagged by human) tags of global news
- has some main topics (5-10), e.g. tech, business, sport, ...
- has to be relatively big (10000+ tags)
- has be up to date
- it would be nice that the keywords have frequency weights.
I thought that couple of the biggest news portals has tag cloud which (partly) fits on the criterions. But I didn't find anything on BBC News, CNN News, Reuters, ...
Interestingly I found some portals in my mother language (hungarian), I cant believe that, there isnt anything on global level.
I dont need API, I can parse the HTML if necessary.
Maybe a corpora can be useful.

Thanks.

[editor]

Have you tried Google N-grams - you can download their data from
http://storage.googleapis.com/books/ngrams/books/datasetsv2.html