Estimate classifier performance on non-Reddit data

TromboneDavies / PolarOps

0 stars 0 forks source link

Estimate classifier performance on non-Reddit data #46

Open divilian opened 3 years ago

divilian commented 3 years ago

As a first cut, run the classifier on some tweets that are obviously polar/non-polar, some NYT comments, some Tumblr posts, etc., and see whether it passes sanity check. Eventually, we should do this in a principled way: hand-tag a set of Twitter/NYT/Tumblr data like we did before, and get actual accuracy numbers for comparison.

divilian commented 3 years ago

Don't start this until #44 is finished.