We should write a plugin which normalises headlines, essentially a de-click-bater. By training a NN on neutral headlines vs doom/anxiety-inducing/clickbait headlines we can essentially provide the inverse as well.
I actually wrote a half baked version of this to replace the headline with a machine learning generated summary of the article. Headlines wound up being sort of long and clunky but I did like them a lot more than the sort of headlines being discussed here (the reason I wrote it in the first place). Maybe I should finish that.
