Ask HN: Best real day-to-day language data sets?

Hi there, I've used Cicero's work to run some NLP code against in the past so I could practice some latin. I am now looking for something more mundane, which reflects real day-to-day usage of languages.

I am particularly interested in data sets for german, but I think I can use Folha de Sao Paulo's texts for brazilian portuguese (it's one of the biggest newspaper and media portal in Brazil and has a respected language manual).

Any suggestions for other languages? Scraping data sets would be OK!

PS1: I don't consider NLTK-level data sets really useful here

PS2: for english I suppose the whole internet and NLP frameworks are enough already

PS3: data sets don't need to be tagged in fact, anything ressemblying a corpus will suffice as it would be used for X-as-second-language :-)