Emerging Theme: Social Media and Crisis Mapping

LESTER HOLT: By air and by boat, the rescuers continue to come. Harris County Constable, Alan Rosen, and his deputies are among many answering the urgent pleas of those waiting for aid. We joined them today as his fleet of high water rescue trucks rolled out looking for those in need. And while the size of the trucks are impressive, they can only go so far.

ALAN ROSEN: We did have one get stuck yesterday with some very high water, seven feet. We had to deploy boats.

LESTER HOLT: Rosen is relying on an unlikely source to direct his teams where to go. This is one of the biggest natural disasters of the social media age. How has that affected how you are dispatched?

ALAN ROSEN: Social media has helped us message, find out where people needed to be rescued. People were on their roofs. It helped us deploy assets.

LESTER HOLT: #SOSHarvey has been trending on Twitter since the flooding began-- users tweeting out addresses where people needed to be evacuated. "2 adults 2 children water at waist level 9015 Sandpiper #HarveySOS". Facebook has a safety check where users can alert family members to their whereabouts. Those still in need of help are urged to post their addresses and phone numbers and sit tight. Help will come.

SURVIVOR: We're going to be OK, I think.

ALAN ROSEN: We're still in a search and rescue mode. And we're going to continue to focus on saving lives until the water recedes and everybody is safe to go home and clean up their lives.

LESTER HOLT: And, as you saw, the constable working right alongside with the National Guard and their high water vehicles on that rescue. Just some of the heroism we're seeing here in Texas as a terrible situation brings out an incredible showing of humanity.

Hey NBC News fans, thanks for checking out our YouTube channel. Subscribe by clicking on that button down here. And click on any of the videos over here to watch the latest interviews, show highlights, and digital exclusives. Thanks for watching.

Patrick Meier: How can social media help in disaster response

PATRICK MEIER: I work in a number of different networks and projects in social media and disaster response. One of them is a Digital Humanitarian Network, which is a global network of technology savvy, digital volunteers who basically make sense of the big data that gets generated during disasters on behalf of established, traditional humanitarian organizations because these organizations have no idea how to make sense of this big data. It's a very new world for them.

Typically, well, like the Nepal earthquake-- happening within a few hours of the Nepal earthquake, the Digital Humanitarian Network was activated by the United Nations to carry out a number of different missions. One of them was to look through social media and find urgent messages. So what we did is we used a new platform called micro macros, which crowdsources the filtering of social media information to quickly find those messages that have to do with urgent calls for help.

So what you do is you give the algorithm a lot of examples of what urgent tweets looks like. And the algorithm starts to learn what the similarities are. So if you say, OK, these 100 tweets here are all about urgent needs and you send it to the algorithm, what the algorithm does is it looks at the similarities between all these 100 tweets. And it says, OK, now I understand why all these tweets are similar. I'm going to find more tweets like these 100 tweets. So it's that simple.

SensePlace2: Visual Analytics and Big Data for Spatiotemporal Sensemaking

[MUSIC PLAYING]

JOSHUA STEVENS: By the time I finish this sentence, more than 500 million users on Twitter will have sent over 37,000 tweets. By the end of the day, these users will send more than 400 million tweets. There is clearly a lot of conversation happening on social media. But what can it actually tell us?

Hi, I'm Joshua Stevens, a 2nd-year PhD candidate in Geography at Penn State's GeoVISTA Center. As part of the IGERT in Big Data Social Science, I'm working on a project that attempts to make sense out of this torrent of tweets.

Humans are very social, and we are now more connected to each other and various sources of information than ever before. We enjoy talking about the events happening around us, sharing our successes, lamenting our failures, and distributing information in times of crisis. The Internet and social media allow us to do this on a global scale. This means that where people are talking about is just as important as what people are talking about.

Now, you may be thinking, doesn't Twitter already record my location and include it with my tweets? It certainly can. But less than 1% of all tweets include geolocation data. So the more difficult question and the concept we're interested in is, how do we determine the geographic context of the other 99% of tweets? And more importantly, how can we determine the locations people are talking about, not just the locations they're tweeting from?

To do this, we've created a tool called SensePlace2. SensePlace2 works by analyzing tweets to identify topics and locations mentioned in the text. We use sophisticated entity extraction and geolocation algorithms to map the places mentioned in each tweet, pairing mentioned locations with both a tweet and a timestamp. When this is done for hundreds of millions of tweets, there's simply too much information to take in all at once. This is where advanced cartography and interactive features, the key elements of geovisual analytics, come into play.

Let's take a look at SensePlace2 in action to see how this works. We'll run a search using the term "protests." You might reason that some places will mention protests more than others. But where are mentions of protests more common? And which locations tend to get mentioned with protests at the same time? SensePlace2 can help us explore these questions.

The Map View shows all the locations that are mentioned in tweets containing our search term. A Place-tree view and word cloud show the most frequent place mentions in our search results. The dark red areas and the timeline reveal that tweets mentioning protests were more frequent earlier in the year. We'll constrain our search to this time period by adjusting the Temporal Controls. Notice that the top locations are different now.

With SensePlace2, users can explore the geographic landscape of tweets in an intuitive, interactive way. By enabling analysts to understand the what, when, and where of social media, our research demonstrates how geovisual analytics can support spatial understanding and the interpretation of enormous and complex data sets. This enriches our ability to ask important questions about the topics that affect people and places and how these relationships change over time. On behalf of the SensePlace2 Team, thank you for viewing our video.