SLJ-Who is a silly command line tool to guess who is more likely to be the author of your favorite parody twitter account.
In Jan 2014, I said this:
Then @endtwist in Jan 2015 actually did the work:
And @techcrunch later here:
I applied a similiar approach using TF*IDF and also cleaned the data a bit before classification. Changes results a bit.
Go to https://apps.twitter.com/app/new Create an application.
Under "Keys and Access Tokens", grab your Consumer Secret & Consumer Key.
gem install bundler
git clone https://github.com/aaronbatalion/slj-who
cd slj-who && bundle install
# setup with your twitter key/secret
/slj-who.rb config --key=THIS-IS-YOUR-KEY --secret=THIS-IS-YOUR-SECRET
# dont forget to set your key/secret
./slj-who.rb run --authors=dcurtis,abatalion,levie --parody=startupljackson
Produces output like:
Who is startupljackson? Comparing 3 accounts
================================================================================
[dcurtis] fetching 1000 tweets: .............
[dcurtis] creating term frequency doc
[abatalion] fetching 1000 tweets: .................
[abatalion] creating term frequency doc
[levie] fetching 1000 tweets: ......
[levie] creating term frequency doc
[startupljackson] fetching 1000 tweets: ..................
[startupljackson] creating term frequency doc
building a similarity matrix from term-freq docs
[48.60%] levie
[47.04%] abatalion
[45.65%] dcurtis
lots... feel free to fork.