Using Telecom Data to Verify Social Media Posts

Your next video will start in

Recommended Videos

  • Info

  • Comments


Nov. 11 (Bloomberg) -- Buzzient CEO Timothy Jones discusses sifting through social media during a crisis with Cory Johnson on Bloomberg Television's "Bloomberg West." (Source: Bloomberg)

Important to spend time assessing these tweets and facebook posts when probably everyone needs some help?


First of all, condolences out to all of the folks who have been affected by this tragedy.

And i wanted to say also, on veterans day i think it's appropriate to recognize the fact that the united states marine corps is already onsite trying to help out some of the folks.

To the point of checking the veracity of some of the data, one of the things is the signal-to-noise ratio program.

In capturing all of this information, one of the things we always advocate is having ways to automatically and intelligently filter the information so you can get true, solid signals from the torrent of information that's probably flooding in right now.

So how do you do that?

On some level gest i guess that's what we all do with all information.

But how did you figure out what matters and especially in social media, what's true?

That's a great question.

So one of the things you've seen a lot of organizations start with is what we refer to as the social media command center approach.

Or what i refer to as the social media intern approach where you basically have people stare at screens and they try to manually assess the validity or tweets or the value of tweets.

One of the things we learned years ago and a lot of this was driven by one of our largest customers, was that that doesn't scale.

When you get to environments where you have tens of thousands, maybe hundreds of thousands or millions of tweet, you need ways to automate.

That there's a whole range of technologies in a domain which is referred to as sentiment anal sills.

Whereby you can use machine learning approaches to automatically score or evaluate the veracity or the relevance of various tweets or facebook postings.

And -- i'm going to interrupt you.

I want to talk about that.

Sentiment analysis.

Machine learn something basically sort of a-b testing, right?

If it's this, it's that.

If it quacks it's it's a duck, if it barks it's a dog.

Yeah, yeah.

One of the things we're very, very clear about is there's a huge difference between what i'd like to refer to as let's say current state-of-the-art machine learning and artificial intelligence.

We're not talking about trying to replicate how the human brain work.

But rather there are ways that you can use methods today to very quickly assess the significance or relevance of a tweet or posting to a particular subject, so, hey, corey mentions quack and corey, is you know, a bird hunter or a bird lover so this is a relevant tweet.

And also tonality.

Did he say something positive or negative about a duck quacking.

That's what we mean about that.

By automating the collection of -- the collection of tweets using sentiment analysis tools today, you can reduce the -- in many ways reduce the noise and hone in on the true signals and then you allow human beings to actually determine whether or not the information can be acted upon.

So that's what we mean about using sentiment analysis tools.

Along with human beings in the process.

I've got to imagine that the location data and g.p.s. data that's incorporated in so many tweets and posts has got to be a real big help in this situation.


These have come a long way.

I would -- let me put it this way.

I think there have been huge advances in the level of geo-location information which is attached.

What we're cautious about is a garbage-in, garbage-out program.

I can go into facebook and twitter and put a location.

I can say i'm in pars and thenky actually tweet from another location.

Now, my tweet may actually show up through the a.p.i. of twitter as being in the actual remote location but my profile will say paris.

So where am i really?

There are a number of issues still to be resolved around geolocation.

This is where we've been working more closely with telecom producers -- providers because at the end of the day, if i'm using twitter or facebook on a mobile channel, the best information about where i really am is actually on that phone.

And one of the things we expect to see in the future are stronger, let's call them social e-911 services made available by the telco carriers so they can very accurately identify the location of someone who is on social media.

Interesting stuff and a horrible story.

Thank you very much.

My pleasure.

This text has been automatically generated. It may not be 100% accurate.


BTV Channel Finder


ZIP is required for U.S. locations

Bloomberg Television in   change