Deep Learning on Title and Content Features to Tackle Clickbait

#1

Post by abhisvnit » Mon, Mar 13, 2017, 11:19 AM UTC

Deep Learning on Title and Content Features to Tackle Clickbait
linkedin.com

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#2

Post by abhisvnit » Mon, Mar 13, 2017, 12:45 PM UTC

code available here: https://github.com/abhishekkrthakur/clickbaits_revisited

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#3

Post by jj12345 » Mon, Mar 13, 2017, 12:47 PM UTC

Thanks for the nice, condensed article. Generating features from BeautifulSoup isn't something that I've considered before.

I'm still going through Yoshua Bengio's new book on DL, but if anyone is free to comment: what are the justifications for the general architecture? Why use LSTMs with the Glove embeddings?

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#4

Post by volker48 » Mon, Mar 13, 2017, 12:51 PM UTC

jj12345 wrote: post #3

Thanks for the nice, condensed article. Generating features from BeautifulSoup isn't something that I've considered before. I'm still going through Yoshua Bengio's new book on DL, but if anyone is free to comment: what are the justifications for the general architecture? Why use LSTMs with the Glove embeddings?

Seems like everyone uses the glove embeddings for any text based DL project.

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#5

Post by joosters » Mon, Mar 13, 2017, 2:21 PM UTC

Simple filter for tech articles:

$clickbait = /Deep/;

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#6

Post by baxuz » Mon, Mar 13, 2017, 2:30 PM UTC

You could add any title that's formulated as an imperative. "You won't believe..." "Guess which..." "You should..."

Also titles that are formulated as a simple subject - predicate - object sentence: "XY considered anti-pattern" "Trump is right" "Hitler did nothing wrong" "Drunk girl shows tits" "Homeopathy is the future of medicine"

Same works if formulated as a question: "Is Trump right?" "Has Hitler done nothing wrong?" "Is homeopathy the future of medicine?"

Bonus points for exclamation marks, pound signs and uppercase words.

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#7

Post by minimaxir » Mon, Mar 13, 2017, 2:41 PM UTC

baxuz wrote: post #6

You could add any title that's formulated as an imperative. "You won't believe..." "Guess which..." "You should..." Also titles that are formulated as a simple subject - predicate - object sentence: "XY considered anti-pattern" "Trump is right" "Hitler did nothing wrong" "Drunk girl shows tits" "Homeopathy is the future of medicine" Same works if formulated as a question: "Is Trump right?" "Has Hitler done nothing wr…

I wrote the original article visualizing clickbait from scraped Facebook data: http://minimaxir.com/2016/08/clickbait-cluster/

Yes, there are obvious tropes of clickbait. Facebook, however, is cracking down on them, so there's been a slight brinksmanship between "how do I get people to click articles without following the tropes?"

From the visualization in my article, you can see there is a spatial blend between sources like the NYT and BuzzFeed when subjects like kids and Pokemon are brought up.

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#8

Post by grabcocque » Mon, Mar 13, 2017, 2:42 PM UTC

The problem is, we don't have a clear definition of what clickbait is.

nouninformal (on the Internet) content whose main purpose is to attract attention and encourage visitors to click on a link to a particular web page.

But that's basically everything on the web.

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#9

Post by abhisvnit » Mon, Mar 13, 2017, 2:47 PM UTC

baxuz wrote: post #6

You could add any title that's formulated as an imperative. "You won't believe..." "Guess which..." "You should..." Also titles that are formulated as a simple subject - predicate - object sentence: "XY considered anti-pattern" "Trump is right" "Hitler did nothing wrong" "Drunk girl shows tits" "Homeopathy is the future of medicine" Same works if formulated as a question: "Is Trump right?" "Has Hitler done nothing wr…

The point that the article is trying to explain is clickbaits cant just be classified only by using these titles. The content of the webpages also plays a big role :)

Re: Deep Learning on Title and Content Features to Tackle Clickbait

#10

Post by mtgx » Mon, Mar 13, 2017, 2:52 PM UTC

baxuz wrote: post #6

You could add any title that's formulated as an imperative. "You won't believe..." "Guess which..." "You should..." Also titles that are formulated as a simple subject - predicate - object sentence: "XY considered anti-pattern" "Trump is right" "Hitler did nothing wrong" "Drunk girl shows tits" "Homeopathy is the future of medicine" Same works if formulated as a question: "Is Trump right?" "Has Hitler done nothing wr…

Not all clickbait headlines are written like that.

For instance: "Russia hacked US power grid" doesn't have any of those, and yet it was a completely clickbait/sensationalist/borderline fake news headline from WashPost. How is AI going to deal with those?

https://theintercept.com/2016/12/31/russia-hysteria-infects-...