> classify user generated content as political or not is non-trivial.
There's probably a set of 50 or so keyword combinations, plus the names of current and former politicians and common misspellings, that would cover 90% of cases.
There's probably a set of 50 or so keyword combinations, plus the names of current and former politicians and common misspellings, that would cover 90% of cases.
- I refuse to pay this "Bill".
- That looks like an Obama suit.
Are these political statements? It's naive to make such a claim that a whitelist of word is sufficient. The "facebook is censoring me" gang will eat you up in a jiffy.
Even extremely complex models will fail this test.
List B: obama, pelosi, clinton, trump, pence, mitch, reagan, bush, maga, antifa, president, congress, representative, democrat, republican
If a comment has a word from list A and list B, it's political.
Using this simple scheme on Facebook would be something I would do if possible.
I wonder if a general black list exists for various websites already.