I mean I fucking hate when the NYT does things like publishing that anonymous official’s “resistance” bullshit open. But it is extremely weird trying to divine qualitative conclusions about “obsession” and unfair negativity exclusively from contextless word frequencies

Also it’s kind of silly to do all the work of scraping the full text of 40 years of articles and not do some basic NLP so that you don’t have to rely on raw word counts. I mean, the author is aware that “Trump” is not just the name of a president, but a worldwide brand?

For example, this 1000-word story about Clinton's emails mentions "Clinton" 26 times

https://www.nytimes.com/2016/05/26/us/politics/hillary-clinton-emails-campaign-trust.html

This 1,560-word story about Trump's real estate mentions "Trump" 52 times. At least 20 of those times are part of a brand name, like "Trump SoHo" and Trump Tower...

...I think it’d be strange to then argue the NYT here is more “obsessed”/ “negative” toward Trump than Clinton based on a naive word count. Again, i agree almost any metric will have Trump over Clinton. But why even do the kind of counting that Google avoided back in 1998?

oh god, I remember this creative use of small multiples mixed with arbitrary bullshit. I’m sorry I bothered to make a Twitter thread, but @cjr should honestly know better.

h/t @colourmeamused_