fabian flöck.

accessible data science tools:
interactive tools to explore and analyse digital traces.

mediamonitoring platform.

The GESIS Mediamonitoring platform allows (i) point-and-click data exploration of social media data from three platforms, and (ii) powerful API/code access through binder-ready Jupyter notebooks.

interactive notebooks.

Several Jupyter Notebooks that are made interactive through "app mode" to create low-threshold data analysis tools for researchers not familiar with code.
Try this example (more will be added)

WhoColor (alpha).

WhoColor creates an on-demand color-markup of the original authors of the text of any article on the (english) wikipedia.

whoVIS (alpha).

whoVIS is a graph visualization of editor disagreements in articles over time, based on the accurate authorship mining of WikiWho.

preface: we wrote a paper about applying whoVIS and WhoColor (and Contropedia) to "Gamergate Controversy". In there you can see some examples of what type of insights one can gain with those tools. it was was presented at the "Wikipedia, a Social Pedia" workshop at ICWSM'15.

developed @

WhoColor. (under development)

what it does: WhoColor is a (.js) userscript for the Tamper-/Greasemonkey browser extensions for Chrome and Firefox. when you open an (english) wikipedia article it creates a color-markup on the text, showing the original authors of the content, an author list ordered by percentages of the article written and (soon) additional provenance information. it also has the ability to show conflict about certain words and the adding/deleting history of a given word (these features are currently being added and might already work when you try it out).

how to install:
- step 1: -
a) for Google Chrome: download and install the Tampermonkey extension.
b) for Mozilla Firefox: download and install the Greasemonkey extension.
- step 2: -
once you have installed one of the *monkey extensions successfully, go to whocolor.user.js.
your *monkey browser extension should then automatically give you an installation prompt that you need to confirm.
if that doesn't work, go into your *monkey extension's menu and look for the option to add a new userscript. copy the content of the WhoColor script and paste in the window for the new *monkey script. note that this script (like the server-side infrastructure) is currently being developed and can change daily. so check back frequently to update.

If your installation was successful, you should see the script appearing in the list of installed userscripts inside the *monkey extension.

how to use:
1. go to any wikipedia article page on en.wikipedia.org. you can also go to an older revision of a page via the revision history.
2. you should now see, and be able to click on, the "WhoColor" tab to the left of the default "Read" tab in the head navigation of the article.
3. once you clicked on the WhoColor tab, you might have to wait until the necessary authorship information is loaded. this can take a while for articles with a lot of revisions and changes, especially if it is the first request for that article (we're working on the speed).
4. after the loading is done, you should see a list of authors to the right, ordered by decreasing percentages of how many words (+special characters) in the currently shown article they have originally written.
5. click either on an author to enable mark-up of all their words or click on a word in the the text to see the author and all other words written by them.
6. on mouseover you will see an outline of all words written by the same author.
7. you can deselect authors and their words by clicking on them again. the number of simultaneously selectable authors is limited, as the number of easily-distinguishable colors is limited.

Troubleshooting: check if your browser permitted the script to be executed and establish all connections and if not, allow it to run. (e.g., Chrome would show you a small shield symbol in the address bar when indicating that it blocked a script)

current issues (we're on it):
- there are currently still a couple of highlighting errors, where certain parts of the text are not marked, although the WikiWho api is delivering (correct) provenance information for them. these are mostly template-related issues. for example, templates on top and bottom of the article, but also references are currently not always marked with their authors. this also includes infoboxes and nav-templates, which we are aiming to mark up as well.
- an author on the right-hand list might not have visible text in the article. this can happen a) because of the bug described above or more likely b) because the percentages of authorship are calculated on the backend wikimarkup-text and certain functional characters are simply not showing in the front-end.
- loading errors / long loading times: loading can take a bit longer (sometimes) or fail completely (rare). this will happen for articles with a lot of revisions and changes, especially if it is the first request for that article. so you *can* try "George W. Bush", but i'll give no guarantees.

features under development:
We are currently implementing these features and the are either already available or should become available soon.
- conflict view: a view that colors those words in the article in a stronger tone of red the more deletes and reintroductions (hence, arguably: disputes) they were subjected to in the past. to (de)activate, click on the two speech bubbles in the upper right corner: 1
- age view: highlights the "youngest" words in an article, the more recent the additions, the more salient the highlighting
- word history dialog: enables marking up a sequence of words with the mouse to show and inspect the periods of time when the selected words were present (blue background) in the article or when they were absent (white background). it also shows who removed the content, for how long, and which user reintroduced it. it can hence aid in understanding who the antagonists were inpossible disputes indicated by the “conflict view”. it is available in all views by marking text with the mouse and then clicking on the bar that appears on the bottom of the page.

planned features:
- show for each word in the standard provenance view how old it is and in which revision exactly it was first introduced, with a link to the diff.
- more languages. in principle not a problem, but you know, has to be tested and so on.. high on the priority list.
- "recently touched" view: will highlight any word that was recently changed (with definable time-span).
- more author information: show authors' total edits, functional roles, etc. in an overlay window.

source code:
available on github .

we briefly introduced WhoColor in an ICWSM'15 workshop paper: "Towards better visual tools for exploring Wikipedia article development – the Use Case of 'Gamergate Controversy'"

WhoColor was built by Felix Stadthaus, as a bachelor thesis under supervision of Fabian Flöck. it is heavily inspired by and extends on the work done for “WikiPraise” by Wikipedia user NetAction, which was based on the now-defunct Wikitrust api.

behind the scenes:
a server-side WhoColor service retrieves the requested article and revision from the userscript and queries the WikiWho api for data about the provenance information of each individual token in the text (for the revision of the article the user is looking at), based on the WikiWho algorithm. it then produces the wiki-text with color markup and feeds it back to the userscript.

1 https://commons.wikimedia.org/w/index.php?title=File:Speechbubbles_icon.svg&oldid=148758659 // By MGalloway (WMF) (Own work) [CC BY-SA 3.0 (http://creativecommons.org/licenses/by-sa/3.0)], via Wikimedia Commons

whoVIS. (alpha version)

we developed whoVIS, a prototype of an editor-editor interaction network visualization for individual articles, based on the word/tokens deleted and reintroduced by editors. try it out here. be sure to read the "how to" section. it's in an early stage and pretty slow when loading up, so have patience when you try it. we are currently adding more and more articles. the WWW'15 conference demo paper (was presented May 2015) with more details about the network construction etc. is available here.