I've added a little script to my TroveTools Python library that takes a newspapers search query and then extracts for each year both the number of matching articles and the proportion this represents of the total number of articles.
The results are saved as JSON allowing them to be easily graphed using tools like jqPlot
Here's an example, a search for articles containing any of the following keywords: 'chinese', 'chinaman', 'chinamen'. (The historian of Chinese Australia in my house successfully predicted the 1888 peak!)
Or what about 'floods':
I put out a call on Twitter for suggested searches and I've been adding them to a Trove Experiments Flickr set.
I might do a few more, so feel free to add some requests to this thread and I'll try and run them.
I'll be adding a blog post with more details soon to discontents.