Twiring Diagrams – Data Visualization (Art) from Tweeting Behaviourhttps://twiring.
ca/ShahamBlockedUnblockFollowFollowingApr 29Twiring comes from Twitter+Wiring — because these diagrams look like ‘wiring’ diagrams.
but also, twiring is to To glance shyly or slyly and that’s precisely the functionality of this form of visualization.
The InspirationI was Inspired by Jim Vallandingham’s Sentence Drawings— which were originally inspired by Stefanie Posavec’s original art of Sentence Drawings — and thought of doing a similar visualization for a user’s tweeting behaviour.
Some of my old projects had looked at generating Twitter Ego-Networks and then algorithmically extracting and depicting topics from tweets so it was a little bit of a mash-up.
Posavec’s original Sentence DrawingsThe concept of Posavec’s sentence drawing is pretty simple: you draw a line relatively equivalent to the length of sentence (word count), make a right turn, draw a line corresponding to the next sentence, and so on and so on.
What this results in is a visual that is congested where it encounters continuous short sentences and open where a long sentence appears.
Posavec’s original Legend and Basic StructureTwiringTwiring takes the above basic structure of Sentence Drawings, replaces the sentence with a tweet from a user and performs a conceptual twist: instead of word count of the tweet, the length of a line represents the time difference since the last tweet.
The spatiality of a Sentence Drawing becomes a way to visualize the time dimension of a user’s tweeting behaviour.
Along with this main switch, I add some other visualization codes to add more information on the drawing.
I also added some circles in the background of the drawing to indicate the popularity of a tweet.
Furthermore, I included a search functionality (basic string matching) to filter the lines/tweets on the drawing — I thought this was cool since you can see which terms appear when and narrow down on tweeting behaviour around a specific term.
Visualization SummaryEach line is a tweetBasic structure: at the end of each line, you make a right turn and start drawing the next lineThe black box with a + inside is the oldest tweet, the empty black box is the most recent tweetThe length of a line is relative to how long it has been since the last tweet by the userThe colour of each line is the topic the tweet has been grouped intoThe thickness of each line is how closely related it is to its topicClicking on any of the legend items will toggle their visibility on the diagramThe search bar (type and press enter) can also be used to filter based on a very basic comparison of your input with text in tweetReplies and non-replies buttons can toggle tweets respectively.
(note: reply to self is NOT considered a ‘reply’)The circles are drawn at the start of each line indicating popularity (relatively speaking) of tweet based on RTs and LikesHovering over a line shows more details about the tweetThis diagram can be panned, zoomed, and saved using its toolbar on the rightNotesI have hosted it as a web app for you to use and play — be advised the topic modelling and collection of tweets can take some time (0–3min).
Topic modelling has some randomness so a refresh will recalculate them and you will lose current configuration.
I only analyzes a maximum of 1,000 most recent tweets.
I also filter out RTs/Likes or tweets with less than 50 characters.
The topic modelling itself is naive, lightweight, and unfiltered (does not use POS tagging or NLP for pre-processing text) for scalability/speed reasons.
I do not store anyone’s tweets and I only ask for read-only permissions for twitter authentication.
ca/Sample UsageVisual feel of how long breaks you take between tweetsSearch a specific term and see how/when it appears in your historyAnalyze other user’s tweets in the same wayImmediately see popular tweets stand out and spot any patternsand so on and so on.