By GREGORY ZELLER //
In a world where “fake news” proliferates and those daily intelligence briefings really do drag on, there comes an artificially intelligent response.
Entrepreneur Blake Cornell, a self-described “nerd’s nerd” who routinely follows bright ideas with several thousand lines of code, doesn’t necessarily intend Long Island Tech News to be the antidote to agenda-driven phonies masquerading as journalists. But by carefully selecting his news aggregator’s sources – and giving users unprecedented control over their newsfeed’s emotional content – the cofounder of Sayville-based e-solutions provider Web Source Group has created the perfect propaganda filter, and that’s just one of his creation’s breakthrough functions.
At heart, Long Island Tech News is a robotic engine that fetches updated news stories every 15 minutes (672 times per week) from roughly 80 local, state, national and international sources, ranging from news outlets to universities to government agencies. Coded to seek out stories involving technology, business, Long Island or any combination thereof, the engine analyzes, matches and stores them in its searchable, constantly evolving database.
That’s fairly basic stuff: Aggregators like Feedly and Google News already offer customizable feeds, while sites like Metanews have been lining up third-party headlines for two decades. Cornell’s creation breaks ground by focusing on Long Island – “I didn’t want to do national,” he noted, “because if it grew wings, I’d need to plan for that” – and especially through its use of NLP, and no, that’s not a reference to neuro-linguistic programming, a widely discredited pseudoscience favored by carnival hypnotists.
In this case, the world’s “first NLP-based AI tech news engine” spices up its artificial intelligence with natural language processing, a computer-science field focused on the interactions between automatons and human languages – the digital secret sauce in Cornell’s intuitive search system.
“Traditionally, people can only search by keywords,” he told Innovate LI. “Now they can search by emotion.”
Basically, Long Island Tech News measures the “contextually aware sentiment” of an article’s specific language, allowing users to search stories based on their inherent levels of anger, disgust, fear, joy and sadness. After a traditional keyword search field, each of those five emotions gets four boxes – Not Likely, Unlikely, Likely and Very Likely – and users can click some, all or none of them.
Cornell pitches it as a time- and effort-saver in a digital world overrun by fake news, repetitive reporting and other inefficient distractions.
“The engine determines the emotional sentiments and stores them as attributes in each individual news article, which can be searched for later,” he said. “So instead of going to 80 websites, I go to one, and instead of sifting through thousands of articles, I can search via emotion and find the needle in the haystack.”
As an example, the inventor searched Long Island Tech News for “Donald Trump” and clicked several emotion boxes, turning down the anger and fear and turning up the joy. While most “Trump” searches turn up boatloads of vitriol, this customized search of the engine’s 80-something sources produced two results: a story about the “Trump effect’s” positive influence on Japan’s Nikkei Index and what Cornell called an “oddball” return focused on the Green Bay Packers.
“Trump was in there,” he noted. “But in the story, the joy was really for the Packers.”
It’s not an exact science, yet, hence the beta run. But Long Island Tech News’ ambitions are high, and they don’t stop at measuring news-article sentimentality.
The engine – which processed 8,130 articles between its Oct. 7 launch date and 3 p.m. Dec. 28 – also caters to the so-much-info-so-little-time generation with tidy article summaries. While searchers can link directly to source articles, they can also breeze through a summary (5,000-word articles reduced to 500 words) or click a button that reads the Cliff’s Notes version aloud, freeing them to multitask.
“There are junk words in language,” Cornell noted. “And there’s repetition in content. Basically, you can concatenate two sentences together and remove the filler.”
As it does with its sentimentality protocols, Long Island Tech News relies heavily on artificial intelligence for its summaries. Cornell has hired no writers or editors; instead, the engine reads the source stories and writes its own synopses through a combination of protocols borrowed from Watson – IBM’s speech-sensitive AI system – and application-programming interface middleware designed by Cornell.
“I generate no content,” he said. “I have no writers. It’s all completely automated. The system automatically finds the content, tags it, categorizes it and shortens it.
“The whole idea is machine learning, which is all about trial and error,” Cornell added. “You can teach a computer to play Mario on Nintendo, you can teach it to shorten articles.”
While providing a customized newsfeed is Long Island Tech News’ primary function, it offers other potential verticals, according to Cornell, who is also chief technical officer at Garden City-based cybersecurity expert Integris Security LLC.
For instance, organizations can sign up as a news source, have the engine fetch their relevant press releases and then visit the site to see how the releases are playing with audiences.
“You don’t want an angry press release,” Cornell noted. “Corporate institutions can check their press releases to make sure they’re not sending the wrong message.”
The programmer also envisions partnerships with “specific news outlets” that want to provide “extended search capabilities” internally.
Cornell’s extended search capabilities will remain in beta run indefinitely, while he incorporates upgrades including new “trend analysis” functionality – allowing the site to rank its “top” stories – and multilingual support (he’s busily integrating Google Translate’s API).
He’s also looking to improve facial- and object-recognition protocols, helping the engine search source sites’ artwork more thoroughly and, in the process, enhance its own search capabilities.
“As time goes on and it processes more and more images, you can search ‘Chuck Schumer unhappy’ or search terms like that,” he said. “The idea is you can apply the emotion on someone’s face to your search.”
Cornell is even dancing with the idea of being able to predict tomorrow’s news today, by hyper-focusing on analytics and studying trends.
“It sounds out there,” he noted. “But it’s totally feasible.”
While the beta version already includes some advertising “just to test the model,” Cornell has other monetization ideas in mind. He’d first like to focus his news engine on a specific industry “and make it national” – the finance industry is a possible target, he noted – before ultimately licensing out the technology to specific news organizations and helping them incorporate it.
Wherever Long Island Tech News goes next, Cornell – who estimates his startup has cost him about 500 hours of programming and “50 bucks worth of software” – knows his fingers will do the walking.
“I’m the sweat equity guy,” he said. “I secure things. I break things. I develop things. My fingers need to move, man.”
Long Island Tech News
What’s It? Emotional, Island-focused online news aggregator
Brought To You By: Nerd’s nerd Blake Cornell
All In: Fifty bucks and gallons of sweat equity
Status: Finding the joy in the Trump transition