Wish you had a better way to make sense of Twitter during disasters than this?
Type in a keyword like #ChileEarthquake in Twitter’s search box above and you’ll see more tweets than you can possibly read in a day let alone keep up with for more than a few minutes. Wish there way were an easy, free and open source solution? Well you’ve come to the right place. My team and I at QCRI are developing the Artificial Intelligence for Disaster Response (AIDR) platform to do just this. Here’s how it works:
First you login to the AIDR platform using your own Twitter handle (click images below to enlarge):
You’ll then see your collection of tweets (if you already have any). In my case, you’ll see I have three. The first is a collection of English language tweets related to the Chile Earthquake. The second is a collection of Spanish tweets. The third is a collection of more than 3,000,000 tweets related to the missing Malaysia Airlines plane. A preliminary analysis of these tweets is available here.
Lets look more closely at my Chile Earthquake 2014 collection (see below, click to enlarge). I’ve collected about a quarter of a million tweets in the past 30 hours or so. The label “Downloaded tweets (since last re-start)” simply refers to the number of tweets I’ve collected since adding a new keyword or hashtag to my collection. I started the collection yesterday at 5:39am my time (yes, I’m an early bird). Under “Keywords” you’ll see all the hashtags and keywords I’ve used to search for tweets related to the earthquake in Chile. I’ve also specified the geographic region I want to collect tweets from. Don’t worry, you don’t actually have to enter geographic coordinates when you set up your own collection, you simply highlight (on map) the area you’re interested in and AIDR does the rest.
You’ll also note in the above screenshot that I’ve selected to only collect tweets in English, but you can collect all language tweets if you’d like or just a select few. Finally, the Collaborators section simply lists the colleagues I’ve added to my collection. This gives them the ability to add new keywords/hashtags and to download the tweets collected as shown below (click to enlarge). More specifically, collaborators can download the most recent 100,000 tweets (and also share the link with others). The 100K tweet limit is based on Twitter’s Terms of Service (ToS). If collaborators want all the tweets, Twitter’s ToS allows for sharing the TweetIDs for an unlimited number of tweets.
So that’s the AIDR Collector. We also have the AIDR Classifier, which helps you make sense of the tweets you’re collecting (in real-time). That is, your collection of tweets doesn’t stop, it continues growing, and as it does, you can make sense of new tweets as they come in. With the Classifier, you simply teach AIDR to classify tweets into whatever topics you’re interested in, like “Infrastructure Damage”, for example. To get started with the AIDR Classifier, simply return to the “Details” tab of our Chile collection. You’ll note the “Go To Classifier” button on the far right:
Clicking on that button allows you to create a Classifier, say on the topic of disaster damage in general. So you simply create a name for your Classifier, in this case “Disaster Damage” and then create Tags to capture more details with respect to damage-related tweets. For example, one Tag might be, say, “Damage to Transportation Infrastructure.” Another could be “Building Damage.” In any event, once you’ve created your Classifier and corresponding tags, you click Submit and find your way to this page (click to enlarge):
You’ll notice the public link for volunteers. That’s basically the interface you’ll use to teach AIDR. If you want to teach AIDR by yourself, you can certainly do so. You also have the option of “crowdsourcing the teaching” of AIDR. Clicking on the link will take you to the page below.
So, I called my Classifier “Message Contents” which is not particularly insightful; I should have labeled it something like “Humanitarian Information Needs” or something, but bear with me and lets click on that Classifier. This will take you to the following Clicker on MicroMappers:
Now this is not the most awe-inspiring interface you’ve ever seen (at least I hope not); reason being that this is simply our very first version. We’ll be providing different “skins” like the official MicroMappers skin (below) as well as a skin that allows you to upload your own logo, for example. In the meantime, note that AIDR shows every tweet to at least three different volunteers. And only if each of these 3 volunteers agree on how to classify a given tweet does AIDR take that into consideration when learning. In other words, AIDR wants to ensure that humans are really sure about how to classify a tweet before it decides to learn from that lesson. Incidentally, The MicroMappers smartphone app for the iPhone and Android will be available in the next few weeks. But I digress.
As you and/or your volunteers classify tweets based on the Tags you created, AIDR starts to learn—hence the AI (Artificial Intelligence) in AIDR. AIDR begins to recognize that all the tweets you classified as “Infrastructure Damage” are indeed similar. Once you’ve tagged enough tweets, AIDR will decide that it’s time to leave the nest and fly on it’s own. In other words, it will start to auto-classify incoming tweets in real-time. (At present, AIDR can auto-classify some 30,000 tweets per minute; compare this to the peak rate of 16,000 tweets per minute observed during Hurricane Sandy).
Of course, AIDR’s first solo “flights” won’t always go smoothly. But not to worry, AIDR will let you know when it needs a little help. Every tweet that AIDR auto-tags comes with a Confidence level. That is, AIDR will let you know: “I am 80% sure that I correctly classified this tweet”. If AIDR has trouble with a tweet, i.e., if it’s confidence level is 65% or below, the it will send the tweet to you (and/or your volunteers) so it can learn from how you classify that particular tweet. In other words, the more tweets you classify, the more AIDR learns, and the higher AIDR’s confidence levels get. Fun, huh?
To view the results of the machine tagging, simply click on the View/Download tab, as shown below (click to enlarge). The page shows you the latest tweets that have been auto-tagged along with the Tag label and the confidence score. (Yes, this too is the first version of that interface, we’ll make it more user-friendly in the future, not to worry). In any event, you can download the auto-tagged tweets in a CSV file and also share the download link with your colleagues for analysis and so on. At some point in the future, we hope to provide a simple data visualization output page so that you can easily see interesting data trends.
So that’s basically all there is to it. If you want to learn more about how it all works, you might fancy reading this research paper (PDF). In the meantime, I’ll simply add that you can re-use your Classifiers. If (when?) another earthquake strikes Chile, you won’t have to start from scratch. You can auto-tag incoming tweets immediately with the Classifier you already have. Plus, you’ll be able to share your classifiers with your colleagues and partner organizations if you like. In other words, we’re envisaging an “App Store” of Classifiers based on different hazards and different countries. The more we re-use our Classifiers, the more accurate they will become. Everybody wins.
And voila, that is AIDR (at least our first version). If you’d like to test the platform and/or want the tweets from the Chile Earthquake, simply get in touch!
- We’re adapting AIDR so that it can also classify text messages (SMS).
- AIDR Classifiers are language specific. So if you speak Spanish, you can create a classifier to tag all Spanish language tweets/SMS that refer to disaster damage, for example. In other words, AIDR does not only speak English : )
Is social media not used in some cultures (due to cultural sensitivities), and could that effect building tools for disaster management in those countries ?
best, Mike Smith ________________________________
Thanks for reading. Indeed, AIDR seeks to solve a very specific problem, not every problem under the sun. This problem (making sense of social media during disasters) was presented to us by several humanitarian organizations (international and national). So that is the problem we have set out to address/solve with AIDR. We are also modifying AIDR to take streaming text messages (SMS). No humanitarian professionals I know would ever advocate for the use of a single information source to base decisions one. Humanitarians today are “Information DJ’s”, they take multiple tracks of information and make the best mix they can to augment situational awareness. So AIDR simply contributes one track to this mix.
Great stuff Patrick. It is definitly the right approach do deal with large scale disasters. Do not you think that brining in some thematic knowledge, like in the case of Chilean quake, the expected shaking level could help?
That’s what we (Euro-Med Seismological Centre) do to assess the reliability of the crowdsourced info (geolocated pics, testimonies), we also compare them with the results of traffic pattern analysis (flashsourcing) and then are able to detect soem of the errors.
My second point concerns time after the event (you notice that I focus on rapid onset disasters such as earthquakes, you can understand why). Our strategy is to focus on the first seconds to the first couple of hours in our analysis of the data (both crowdsourced and on social networks); In this period, the ratio of eyewitnesses reporting information is the highest and the volume of data to be ingested is lower. Have you explored this aspect, because it may stress on the importance for starting as fast s possible the AIDR analysis.
Keep doing good job!
Thanks for reading and commenting. Yes, you can pull in multiple data sources to cross triangulate and assess reliability. For computational approaches to verifying social media reports, please see:
On your second point, yes, starting as early as possible is certainly important. We’re doing research in this area, for example:
Is this platform adaptable to classify social media streams from other events such as elections or protests? Disaster response suggests it’s limited to mass casualty events or humanitarian crises, but I can easily see how this tool would be useful in political or social contexts.
Hi Ursula, thanks for reading and for your question. Yes, this platform will classify whatever topics that are supported by the existing data. Re elections, have a look at this:
Our Terms of Service requires that Tweet collections on AIDR focus on broad humanitarian related issues.
Pingback: Cómo quitar ruido en el universo digital | Conversaciones con...
Reblogged this on Rashid's Blog.
This looks really promising – congratulations. I have a question about the geofencing you mentioned. How are you addressing the issue that only a very small part of all tweets contain embedded GPS information? In a lot of developing countries you will naturally have only have very few tweets (and no, I don’t want to start the fruitless discussion about representativeness) and by narrowing it down further to only tweets with GPS information you might end up with a very, very small sample that might exclude a lot of relevant information. Any thoughts on that?
Hi Timo, many thanks–as always–for reading, and for your follow up question.
Re problem of small percentage of tweets that contain embedded GPS information:
1) There are computational solutions that can boost this through inference and machine learning;
2) Lack of geo-tagged tweets should be addressed as a shortcoming in policy;
3) Some information is better than no information and one can infer from a small set of data;
4) No one in their right mind would advocate for disaster response based one data source (tweets) alone;
5) Humanitarians are becoming information DJs, trying their best to create the right mix from imperfect soundtracks.
I personally am more interested in fixing #2 than looking for technological fixes. If government/UN agencies want more geo-tagged data, then they should ask for it.
Pingback: Using MicroMappers to Make Sense of UAV Imagery During Disasters | iRevolution
Pingback: Zoomanitarians: Using Citizen Science and Next Generation Satellites to Accelerate Disaster Damage Assessments | iRevolution
Pingback: Foresight at the forefront: Examples in action | Voices from Eurasia - We help build better lives.
Pingback: The Filipino Government’s Official Strategy on Crisis Hashtags | iRevolution
AIDR is the great initiative. I am the student of Disaster & Emergency Management program in York University, Toronto. I would like to testing out AIDR platform. I believe, it will be very useful for my research and data collection.
If you require further information, please contact me at – email@example.com
Pingback: Disaster Tweets Coupled With UAV Imagery Give Responders Valuable Data on Infrastructure Damage | iRevolution
Pingback: A Faster Way to Process Large Amounts of Text Data | Natural Disasters Blog
Pingback: On Being a Digital Humanitarian | Claire Diaz Ortiz
Pingback: Teaching the computer: how we are testing humanitarian machine learning tech with our advocacy partners | the engine room
Pingback: Why Asia needs digital Jedis: Patrick Meier
Pingback: Drones and Artificial Intelligence for Social Good - Quadcopter Blog
I am having problems downloading the tweets.
When I download the CSV file is empty. Is it a known issue?
Should I use another format?
Thank you for your time.
Dear Vittorio, thanks for your note. I left QCRI a year ago and so have not been as involved in the AIDR platform since then. I recommend you contact QCRI directly.
Dear Patrick, thank you for your fast reply and cooperation. I will ask QCRI. I take the chance to tell you that I really appreciated your book “Digital Humanitarians” and the need for technologies that make us more human :).