This week I took my headline gathering script and applied the Ngram and Markov chain library to it. After adjusting the length of the N grams and trying out some “key” words I ended up with something rather simple but effective in terms of this confusion of news and time effect I’m going for. Basically I take all of my headlines and pass them through Generate_from_tokens function to create new headlines with an Ngram length of 6. I liked the results here because the headlines are typically generated but occasionally repeating the source material so when you read through a list of them its unclear whats real and what isn’t. Also some of them worked out to be kinda funny. Here are some of the highlights:
Hey, normals, Martha Stewart wants to Nerve Gas in Syrian man says
The Army could go boring, bad, and really bad
Pepsi pull offensive Kendall Jenner ad after he felt like a ‘clown’
Worlds largest canary discovers Weird, Unexplain benefits sacrificed to cut deficit
Syria and Assad Altered After Sexual Claims Against Beijing: Dalai Lama
Demand For Seal Products From That Viral Second Grade Survey Are Finally Here
Samsung Galaxy S8, Samsung Galaxy Note 8 releases Windows 10
US threatens history
Angry Mets fan tells Alex Rodriguez how he used to revel in his Syrian chemical Attack
‘Other Fish to Fry’: Despite Screaming Person
How Target Botched Its Response to Syria
NBA draft lottery odds: Where have all the MLB superstars gone?
‘Designated Survivor’ Recap: Cato Hosts a Fundraiser With Frederick Douglass
Trump fumbles, Putin as a gay clown
Inside of Job-Killing $1 Billionaires who run Trump’s Israel-Palestine efforts
Rogue Ones best phone ever
Pepsi Pulls Ad Accused of causing index funds, Leon Cooperman is now illegal in Russia connection
RIP Facebook will replace it.
Catastrophic water on growing Uber (or Yahoo or any of it)
Donald Trump Warns Syria, North Korea nuke program, FOI investigation red lines and Kendall Jenner
Bannon Removes Stephen Bannon from Chrome and John Cena had to account?
Controversial Trump Removes Steve Ballmer, Reid Hoffman and NSC: Huge Victory in south Wales game
Toddler attack, US warns of unilateral Syrian gas attack’: Trump Campaigned for break
IPL 2017, Match 1: Yuvraj’s blitz proves to Ban Jehovahs Witnesses as it works to overhaul its business
Exclusive–Sen. Ron Johnson suffers fashion dominant win for the Astros, the reverse
Theranos Found On Shipwreck
What to expect from National Security Council
White Housewives of India
Analysis: For Trump: ‘I now have broken the Oval Office
19 Restaurants Where You Might Just Take A Minute To Appreciate Old Pugs
NBA Mock Draft 4.0: New top-10 picks, new language of gig economy
Trickier-than-usual conditions on the Probes Into Russia’s ‘fanciful’ explanation for Space Project
Rapid rise of clothes moths threatens ‘our own action’ in Syrian gas massacre
Chemical attacked by malicious Wi-Fi networks
Woman Who Degrades Women
haley threatened to enjoy the next year
woman who helped launched an ancient land and sea missiles
dealmaster: today only, hydrogen cars from national security council
van jones: trump remarks on news coverage of gig economy
smile if you think robots can be fatally hacked by pitbull-cross in playground mock draft day
what we do know about that viral security council
youtube tv goes `beyond red lines’
‘he’s taking my place’: trump should be terrified
plan a netflix marathon and we’ll guess what happened
spotify’s new deal with frederick douglass
haley threatens historic fabrics
The exact algorithm changes a bit here but the result is generally the same. I tried adding additional parameters to the markov function and other things but the more complicated I made the process the less interesting the results were to me. I was thinking it would be cool to write out a sentence with the key words and have generated sentences follow them or continue my time and date thing but the key seemed like it needed to be a specific length and I couldn’t figure out how to adjust for that.
Code:
