Why NewsWhip is Building an AI Layer


By   |   October 20th, 2015   |   Reading time: 6 minutes Digital Journalism, NewsWhip Announcements

NewsWhip CEO Paul Quigley outlines why NewsWhip is getting into machine learning, in a big way. 

StarWars

Yes. It’s inevitable. The robots are coming for us.

But there’s no need to stockpile the grenade launchers and tinned beans. At least not yet.

See, robots – or more specifically machine learning / artificial intelligence – is not going to arrive via a robo-tank with lasers and mind control beings. Instead it will arrive in loads of subtle changes in the interfaces you use every day. It will help choose what appears in your Facebook feed, based on what you’ve interacted with before. It will help your phone know what you want to do when you pick it up, and save you pecking through your “favorites” in your phone book. And soon, it will help you at work.

At NewsWhip, we think AI assistance is fast becoming necessary for anyone to navigate and interpret the extremely complex information world we live in, a place where our customers (in media and PR) are at the coalface.

Let me back up a little and show why we should welcome our new robot helpers, especially those of us who work in the thick of the rivers of information that flow online each day.

DancingRobot

See, as humans, we’re in the midst of a complete transition in our relationship with media, from there not being very much of the stuff to it being all consuming.

The last media era was defined by Scarcity (c. 2000 BC – c. 2005):

  • Few people could create and broadcast or publish media (we were illiterate, then few people had printing presses or broadcast towers)
  • Limited information could be found in limited places (a trip to the library to resolve any question)
  • There was not much noise coming our way (we had to seek media, it did not flow to us through push notifications and “feeds”)
  • Because there was less media, our standards were lower (we were OK reading boring crap on the train, better than staring at the wall).

Thirty years ago people thought we had already hit media saturation with “57 channels and nothing on.” Uh-uh. They got nothing on today, the era of Abundance (c. 2005 -> forever):

  • Everyone carries a recording and broadcasting device, and uses it to create and broadcast media (pictures, video, sharing news, text updates)
  • Vast amounts of media and information can be accessed from anywhere (all you need is a phone)
  • It’s hard to get the signals from the noise (there is too much stuff to review)
  • Our standards for what we consume are getting higher, as we have limited time but access to everything (we’re less likely to read the back of a cereal box when we’re bored).

At this moment in time, we’re in transition from habits and technology of the scarcity era to habits and technology for the abundance era.

For professionals, the number of information options is fast becoming overwhelming. For example, picture a trader monitoring prices and information for a basket of 100 stocks she trades via a Bloomberg terminal.

These days, there’s plenty more information happening outside the Bloomberg terminal that’s relevant to the trader. She might also monitor mentions of the stock on Twitter (recently incorporated into the terminal). She might want to monitor the aggregate sentiment in those tweets, or the sentiment for particular groups of tweeters. Track velocity of mentions of different companies, perhaps, and how these differ from a baseline. She might want the same information (mentions, baseline, frequency, sentiment) from LinkedIn. Or from Facebook. She might get online media monitoring (like our software) for mentions of the companies associated with the stocks, and their competitors.

Of course, stock prices might be influenced by political announcements, regulatory enforcement actions, weather, trade agreements, publication of national statistics, changes in legislation, natural resource discoveries, competitor product announcements, conflicts, IP monopoly grants by governments (patents) and a million other things, any of which could be reported first in hundreds of places. So she’ll need to be monitoring all of those variables too. A million pieces of information could be coming in every day, any one of which could be important. But more likely, each piece of information is just more noise. But each piece of noise is a potential signal and there are so many of them – photos, reports, tweets out there.

And that’s just for a trader making buy and sell decisions about the value of a single basket of 100 stocks. What about decisions with even more dependencies and vectors?

Binary

Today, users need help finding the needles in haystacks, the signal in the digital noise.

We’ve argued before for the use of social signals (i.e. the wisdom of the crowd) as an ethical, powerful and necessary method of event detection. If people are chattering about something in your sphere of interest, and you’re responsible for finding stories and disseminating information in that sphere, you need technology that will alert you to that chatter fast.

Currently users of our technology or similar technologies define the keywords, topics, publications, events and people they want to keep an eye on, then our technology monitors hundreds of thousands of news items, pages, Instagram accounts, Twitter accounts, and delivers the important stuff, with a mix of social signals and other triggers of importance.

That’s fine, in the short term. But not for the future.

In the long run, relying on activity from those two billion social network users to ferret out the important stuff simply isn’t enough. For one thing, the user has to correctly define all the things they want to monitor. There’s no fuzziness if they leave out something. If you create a keyword filter for “Ford” it doesn’t automatically add “Mondeo”. Or separate “Apple” (APPL) from “apples” (Granny Smiths). Even if entities can be correctly identified and linked with contextual algorithms, a more personalized system is not yet there.

Meanwhile you don’t want every mention of “Ford”. You only want events and stories of significance. How do you establish that threshold?

If an important enough piece of news breaks about an emissions standard, and you’re monitoring for Ford, isn’t it likely you’d want to be notified of that story too? But how do you define the whole universe of concepts of interest to you up front?

This is where software can help through relationship mapping. A platform should be able to “learn” what stories will be of interest to you, as a trader, as a journalist, as a corporate communicator, based on your activity and the keywords or “training set” of data and interests you provide to it.

Over time it should be able to associate and cluster stories along your sphere of interest (or vectors of interest). Based on your reactions and those of other users, it should establish thresholds for stories you’ll find interesting. And when you do, then – Kapow – it should alert you.

That’s where we’ve started work directionally at NewsWhip. We believe technology should deliver the information people need to them, freeing journalists and communicators to investigate, analyze, create, and allocate resources, without the constant searching and FOMS (fear of missing something).

Machines will do what they have always done better – sift all the information and deliver a result.

The-Theory-Of-Everything

Find your signal in the digital noise with a free Spike trial.

 

Tags: , ,
Still Curious?

Help Center

Blog

Press

About Us

Company

Careers

Contact Us

hello@newswhip.com
+1 917 410 6836

EU Office
47 Merrion Square, Dublin 2, Ireland

US Office
25 Broadway, New York, NY 10004, USA