I'm looking for a dataset of documents (bag of words style is fine) which include time stamps.
I also need it to be semi-coherent, for instance a years worth of news articles which all focus on the economy/finance, broken down by day, or something like that. If it's just a broad collection of documents that aren't really related that doesn't work for me. Something around 30k to 100k documents would be ideal.
I have scoured the internet but haven't really found any datasets of this type besides the NSF abstracts corpus, and unfortunately that doesn't really contain enough in the way of changing temporal dynamics. I need something that's either faster changing (like news), or spans a much larger time interval.