Kirn and Taibbi have fun with a bizarre “sciency” article written by Brookings Inst. The article uses superfancy fake “vector” methods to analyze podcasts by Unpersons and finds “scientifically” that Unpersons are Unpersons. Which they already knew.
I’m a bit more mathy than K and T, so I looked at the actual article to check the method. It’s totally inappropriate for any type of language analysis.
The first strategy used to identify when specific claims fact-checked as false appeared in podcast transcripts was a text similarity measure called the “cosine similarity.” The cosine similarity is a measurement technique that calculates the relationship between two documents (or text snippets) by encoding each document as a vector, with each word of the document representing a coordinate in the vector. The cosine similarity is then calculated using the cosine of the angle between the two vectors. The resulting output is a score from zero to one, with a higher number representing texts that are more similar.
You can’t compare FACTS by correlating the WORDS. This tells you only that the two documents have a similar vocabulary or style. Doing a dot product (direction cosine) on the two sets of words is just an alternate way of doing a correlation. Probably faster to calculate.
Conservatives do tend to have a similar style. For many years conservative writers followed Buckley’s weird grammarrhoid tendencies, in and upon which, at the incipience of each utterance, one shall, as it were, pile every conceivable preposition, and in and upon which, one shall implement and utilize each, shall one say, leaf and papyrus, of one’s Thesaurus upon maximal advantageousness, and in and upon which one shall substitute ‘one’ for all otherwise comprehensible pronouns.
In more recent times each partisan team has its own recognizable names for the other team, like Drumpf vs Brandon.
Matt and Walter aren’t mathy but they have massive common sense, and they instantly spotted the fake math. BUT: The article has so many layers of fakeness that they weren’t able to peel the whole onion. They MISSED the deepest fakeness.
= = = = =
CENSORSHIP IS ABOUT WHO, NOT WHAT.
CENSORSHIP IS ABOUT CASTE, NOT CONTENT.
Censors know WHO they want to censor. Facts are utterly irrelevant to a censor.
Every utterance by an Incorrect Person is a CRIME.
Every utterance by a Correct Person is a COMMAND.
= = = = =
Orwell took great pains to clarify the WHOness of tyranny. Americans consistently miss this part, focusing on WHATness. Winston wasn’t forced to BELIEVE that 2+2=5. He was forced to OBEY every COMMAND by O’Brien, whether true or false or irrelevant.
Brits are more honest about caste and aristocracy. Americans are accustomed to fakely subbing ideology and “measurable achievement” for caste.
