Uses Labeled Latent Dirichlet Allocation (LLDA) to attempt to classify news from two merged datasets as "real" or "fake" based on either article title or article title + hostname
0
stars
0
forks
source link
Test with prepending/appending hostname (i.e. NYTimes) to article title. #2
Turns out, it doesn't actually make that huge of a difference to prepend hostnames.
The testing accuracy went from, like, 82% to like 85%