Visualising what people on Twitter like

08.09.11 Max Novakovic

The Twitter streaming API allows access to tweets on Twitter as they are created. I decided to experiment by writing a data-mining script that collects geo-located tweets from the streaming API that contain people saying that they 'like' something. For this post I have created two visualisations that aim to make sense of the data collected so far and display it in an easily digestible way.

The script has been running for about 5 days in total and has collected 1873 tweets; It has parsed a lot more tweets than that but a tweet is only added when it meets a certain criteria; which is that it is geo-located from the UK and contains a phrase such as 'I like', 'I love', 'I am fond of', etc...

From the data the second most 'liked' thing is Twitter itself which is un-surprising; 'My Life' is 4th and 'My vagina' is 32nd. The most liked things on Twitter are 'That song' and 'This song' which provides a very un-informative insight into the people on Twitter.

The top two images on the right are visualisations of the data. The first image is a 'Tree Map' which was quite simple to put together thanks to Google Chart Tools. The larger the square, the more likes and vice versa. Click here to view the visualisation in full. The data isn't perfect, for example, 'The' is one of the most popular things but I think overall it works quite well at extracting 'likes'.

The second visualisation is a Processing sketch that plots a users location along with what they have said they like. View a larger image of the sketch here. It doesn't really provide any insights geographically but it's a starting point for something that could have a lot more potential for spotting geographical and cultural trends. Download the code and data for the Processing sketch on the Disturb wiki here.

For now, I will leave the spider running for another month or so and then post again with some updated visualisations and hopefully some more interesting data from Twitter.

Tree map of Twitter like data Processing sketch of twitter likes plotted on to a map Pointless filler image




Visualisation of olympics game entries

01.08.11 Max Novakovic

This visualisation shows entries to a nationwide competition which took place over about 5 weeks. The competition consisted of 1,000 stores handing out scratch-cards to customers offering them a chance to play a web based game to win a pair of tickets to the olympics. Please ensure you watch the video in HD (720p)!

The game was widely cross-platform and allowed the game to be played on mobile devices, tablets and desktop computers. In the visualisation circles represent an entry. The diameter of the circle is based on the score the user achieved (a bigger circle is a better score) and the location of the circle represents where the player was located in the country.

The visualisation was created using Processing.

Some numbers:

  • There were 25,372 entries
  • Spread out over 38 days
  • Which is 676 entries a day
  • Or about 1 entry every 2 minutes
  • Time in the video is sped up 36,480 times




OFFF 2011 Barcelona - Let's feed the future workshop

20.06.11 Disturb Media

To follow up on my previous post about 'Aaron Koblin: Let's feed the future workshop' at OFF, I'll talk about some of the projects created in the workshop. It was held by Aaron Koblin, Ricardo Cabello AKA Mr. Doob, Filip Visnjic and Eduard Prats Molner.

Unfortunately, the elusive Mr. Doob was nowhere to be seen on stage during the talk which was a shame. The workshop consisted of 10 applicants who had six and a half hours to create something cool. I think the time restraints affected the output of the workshop a little bit and as a result we saw 3 or 4 completed projects presented by four of the applicants. These consisted of some work using the Kinect and Cinder to route the Kinect output to a browser and then modify it using Chrome and WebGL; some work involving compression of images and audio to visualise what content is actually lost during compression and the 'Receipt Racer' a basic racing game re-worked to run on a receipt printer using a playstation controller and a projector.

To start, I will discuss the Kinect and WebGL. This was demoed by Marcin Ignac and from what he said Mr. Doob helped him to get it working. It's nice to see a browser processing Kinect data in real-time and using WebGL, however, I don't think it demonstrates a real life usage of the Kinect whereby users at home could use their Kinect to interact with a webpage. I don't think this would ever be possible without the use of some kind of plugin so it is still impressive to see it working in a web browser. This image posted on twitter by Marcin Ignac (which you can also see on the right below the video) shows a preview of the demo and the image below it shows a grab of the demo during the talk.

The work involving compression of images and audio was an interesting topic and seemed like a good starting point for some more interesting work. Unfortunately, I can't remember who presented it and I don't have any images of it being demoed but I will try to explain what went on.
Firstly, they took an image at full quality and an image that had been compressed; they processed the images to find the differences between the two and then outputted this difference as a separate image which visualised what the compression had actually done. Because there's no images of this, I have tried to re-create this which can be seen in the three images on the right. First is a full-res version of an image, then a compressed version and finally a version where anything missing from the original image is shown in black.
After showing the image version, they used this same concept on audio files, so they compressed an audio file, compared it with the original and then played out the difference so we can listen to what has been stripped out by the compression. It was interesting with the audio because to me, I couldn't hear a difference between the compressed and un-compressed versions.

Finally, what I think was the most interesting project was the Recipt Racer. This was created by Joshua Noble, Martin Fuchs & Philip Whitfield. There is a video of the project being demoed at the top right of this post; you can also watch a clearer video here. Joshua noble also has some photographs of the project on his website. Using a thermal receipt printer as the 'screen', a projector to project the player on the 'screen' and a playstation controller to move the player around the 'screen' they have created an interesting combination of physical and digital mediums to re-hash a simple racer game.
The receipt printer prints out a basic track which curves from left to right and prints out random blocks to avoid. The projector projects a little dot which represents the player or the car onto the track; the receipt printer then prints more track which causes the movement of the car and finally, the player must move his car around the printed track using a playstation controller to avoid the walls and obstacles. Watch the video to get a good idea of how it works. I particularly like the way the projector shows the crashes using pixelated red explosions.

All in all there was some interesting work as an outcome of the workshop. It would be interesting to see what would have come out of it given the attendees more time, but I think that the time limit was an important concept of the workshop.

Marcin Ignac tweet about Kinect + Cinder + WebGL Kinect + Cinder + WebGL Image compression example HQ Image compression example LO Image compression example - Difference




OFFF 2011 Barcelona - Aaron Koblin

16.06.11 Max Novakovic

'Aaron Koblin: Let's feed the future workshop' was probably the talk I was most looking forward to at OFFF. It focussed on a workshop available for a select few where the attendees would 'work together for one day and present their creations'. I will be covering the results of this workshop in another post, but I think Koblin's introduction to his work and himself warranted a separate blog post.

Aaron Koblin is 'an artist specializing in data and digital technologies'; he is also Creative Director of the Data Arts Team at Google. At OFFF before the workshop attendees presented their work, Koblin gave a brief overview of some of his works. Some projects he mentioned were:

The Johnny Cash project is an interactive music video in which each frame is drawn by a different person. A user will get a frame of the music video and they can then use the drawing tool on the website to re-create the frame in whatever style they choose; limited to a black and white palette. The video at the top right of this post is a version of the music video. Since the video is ever changing, it will probably now be different so go to the website and see it for yourself. Another interesting feature of the project is that frames can be categorised and rated so you can view the video differently based on frame rating, abstract frames, realistic frames, etc.

The video for Radiohead's House of Cards (credits) which used lasers to capture surroundings and shapes to produce a music video that didn't have any video in. You can download the source (written in Processing) and watch a making of video at the Google code page. Koblin also explained how because the video was open source, many spin-off versions of the video had been made and uploaded to youtube including a 3D print of Thom Yorke's face, see image to the right.

eCloud is a sculpture/permanent installation housed inside San Jose International Airport. It features hundreds of panes of 'Smart glass' which are turned on and off to simulate weather patterns around the world. There is a display which details what is being displayed on the panels. There is a lot of information about this at http://www.ecloudproject.com/. The photo to the right (by Spencer Lowell) shows the eCloud installed in the airport there is also a video of it in action here. This project was built in Java and was built using Processing as a base.

Others were ROME which is a completely browser-based music video which makes use of webGL to create realtime 3D graphics, videos and interaction to create an immersive experience. As Aaron said, you really need to see it yourself to experience it so visit the site: http://www.ro.me/.
The Wilderness Downtown another browser-based music video that uses google maps to personalise the music video, features a drawing tool to send messages to your past self and makes use of browser windows to frame different parts of the video.
Finally, Ten Thousand Cents which was a project that used Amazon's Mechanical Turk to ask thousands of people to draw a fraction of a one-hundred dollar bill. The result can be seen in the animated gif to the right.

To see more work of Aaron's, visit his portofolio at http://www.aaronkoblin.com/, it's all really impressive work. I like to see that Processing is used in many of his projects. Also, I think that my favourite projects of his are those in association with Chris Milk and Mr. Doob - ROME, The wilderness downton and The Johnny Cash project.

3D print of Thom Yorke's face based on the House of Cards music video Photo of eCloud at San Jose International Airport by Spencer Lowell ROME sharing tool at OFFF Ten Thousand Cents animation




SameAs visualisation

30.03.11 George Profenza

...or how to fit 132 people in a room.

Monday after work, joined by the youngest and oldest of disturb's employees, I went to the SameAs visualisation event. This was the first meetup where everyone that signed up actually got there, thanks to such great speakers.

The talks were brief, but very informative. I was amazed to see the speakers fit content so neatly in a very limited amount of time.

I was impressed by Stefanie Posavec's passion for gathering data and meticulous work. Strangely enough I've managed to see some of her works for the first time at the Pick Me Up exhibition in Somerset House. Coincidentally this event allowed me to learn more about her work.

Alastair Dant showcased very interesting interactive Flash projects build with his team at The Guardian. From World Cup football and balls to custom budget cuts, carbon footprint calculators, afghan war logs and election swingometers, Alaistair quickly covered different techniques in relation to the data portrayed in a very informative talk.

Brock Craft's discussion revolved around his project using the TFL's Cycle scheme API to plot the availability of bikes which part of these scheme through out London. A lot of helpful tips and insights, including details regarding regulations for requesting sensitive data.

The organizers saved the best for last, as Noah Iliinsky delivered a funny, yet very helpful guide on the DOs and DON'Ts in data visualisation. Speaking of DON'Ts, there was an interesting example of a 3D pie chart (from an Apple keynote) which was tilted at a convenient angle so a smaller percentage visually was perceived as larger. Noah was kind enough to a 50% discount code for the book he co-authored: Beautiful Visualization. The code is valid at the time of this post, but I am not sure for how long. The code is: AUTHD

Unfortunately I am not able to go into great detail about each of the talks, but do check out the event website for more details on the speakers and also keep an eye out. Some sessions were filmed so I would expect them to go online at some point in the future.

sameAs visualisation header full house Alastair Dant Alastair Dant Noah Iliinsky Brock Craft Stefanie Posavec Stefanie Posavec