Here is what it looks like if we just count how many articles are in each topic each year.
There are so many overlapping dots that it isn’t that helpful a graph. It’s a bit clearer when the topics are broken out.
Note a couple of things about this as compared to the graphs back in section 3.1. One is that the Y-axis is much higher. The raw count measure tends to amplify trends. Another is that the graphs are much noisier. There are still enough trends that I don’t think it would really help to add trend-lines, but there are a lot more graphs that look just like scattered points, especially over ten-to-twenty year patches. I think this is basically a flaw in the underlying measure being amplified.
The animation makes things a bit clearer here because the original features so many overlapping dots.