Visualising the edit in a Wikipedia article

10 Oct 2013 - 9:43pm
2 years ago
3 replies
5373 reads
I'm an IEG grantee with the Wikimedia Foundation. I'm working on a project to visualise the edits that have occurred on an article. A blog post about the project.
The history tab in wikipedia does not visually give you much insight into how big the edits were or the time between edits etc. Also when comparing two edits, the changes are show line by line in wikitext. I'm trying to address them by 
  • Making the changes appear like a playback, text & content that have been added appear and disappear as they are removed. (demo
  • A slider to select the edits to investigate. (demo, move the slider to the extreme left to add data ,this is just a prototype with with lots to be done to it)
Some of the challenges I'm facing in building the slider are 
  • Showing the temporal distance between edits while at the same time not showing a lot of blank space and defeating the purpose.
  • Are bars ( I'm currently using bars) the best way to represent edits especially as articles may have thousands of edits.
Any comments or ideas for the slider or the tool in general would be much appreciated.


17 Oct 2013 - 4:06pm
Juan Lanus

Hi cosmiclattes,

IMO the issue stems from trying to graph the time and to use a time-proportional slider. 

Ths updates of an article are not a timed sequence of events, like the heart beats.

They are yes a sequence, in that one happens after the other, actually on top of the previous one.

About the slider, time is continuous while WikipediA updates happen at any time. Trying to map the updated outbursts into a continuous is al the least distorting.

I see the updates as a sequence like 1, 2, 3 ... and might appreciate a pasive timeline tlke thus:

2004  2005  2006  2007  2008  2009  2010 ...

a horizontal line with vertical notches when an update happened, to have a bird's view of the changes history. And a highlighted update, the version currently on screen.

So my version of the control would have a [<prev] and [next>] buttons arrangement, but it wouldn't be that easy.

After thinking a lot about the document versions thing in the past, I realized that I needed to ba able to recognize text blocks, like paragraphs or headings in both versions, to use that blocks as anchors for synchronizing a left-hand and a right-hand version. Even in the case when a document underwent heavy insertions.

One significant issue is that the comparisons should be allowed between any versions pair, not necessarily two consecutive versions. There are two use cases to support this need.

The first is a document that was revised and after the revision had several minot changes, like typos. The useful comparison is between the version previous to the revision, and the last one, skipping the minor changes.

The other use case is a document that was almost fully changed by an editor, maybe for oscure political reasons. Later on the document is restored and the troll version needs to be skipped in comparisons.

Well, these are my thoughts, as an almost completely uninformed participant on your work. What do you think?



18 Oct 2013 - 9:04am
Juan Lanus

Also I forgot to add that it might be useful to try to detect articles that have undergone extensive changes.  

In those cases it might be pointless to try to display a comparison. Displaying both texts side by side would be preferable for the user to read them, over an extremely busy page showing the coincidence of short text spans.  

If you are on Windows it might be useful to visualize the cases with a program line WinMerge, freely available from sourceforge. 

21 Oct 2013 - 12:22am

Hi Juan,

Thanks for your comments. I've been working on a newer prototype since I posted the the question, here is the link

There are two graphs. In the first one ther is no notion of exact time/date, the consecutive edits are shown irrespective of the time taken between them. In the second zoomed in graph the time at which the first and the last edit were made are shown. Also the time between the edits(the bars reprsent the edits ) is proportional to the gap between the bars.

Unlike the usual diff's the text (prev & next ) are not shown side by side instead an animation is show, the user sees words/content appearing and disappearing as they were added/deleted or modified (like a video,

In the current iteration of the tool I'm showing all the changes, even if they are minor typos, but I guess its a good idea to be albe to skip them.

Syndicate content Get the feed