Ian MacKinnon@imackinn · CTO @ Later.com
Digging through the GitHub repo, seems like the extension makes a call to a backed for the deep learning (fair enough). Can you give us a TLDR of how the deep learning goes? Did you train once with some corpus on Theano or Tensorflow, or is there a rules-based system back there?
Michael Joseph@michaelcjoseph · Product Guy. Experimenter.
@imackinn @rahulkapoor90 would love to learn more about this too!
Rahul Kapoor@iamrahulkapoor · Developer by heart but gentle by mind.
@michaelcjoseph @imackinn yes It's a ml program so it learns patterns from different data sets using tensorflow. The dataset consists of about 12,000 headlines half of which are clickbait. The clickbait headlines were fetched from BuzzFeed, NewsWeek, The Times of India and, The Huffington Post. The genuine/non-clickbait headlines were fetched from The Hindu,… See more
freia lobo@freialobo · tech@nyu
@rahulkapoor90 Isn't that in itself inherently completely flawed? Buzzfeed writes plenty of great headlines and WSJ writes plenty of clickbaity ones.
Jason Salas@jasonsalas · Product Director
@rahulkapoor90 @michaelcjoseph @imackinn That's a very clever heuristic. Congrats! I work in product management for online news, so I'm wondering how many false positives your system runs into due to publishers getting creative with their headlines with literacy devices like sarcasm, humor and puns. Thanks! :)