Grants:TPS/Dschwen/Wikimania/2014/Report
Welcome back from Wikimania 2014!
Participant
[edit]Dschwen
commons
Outcome
[edit]- New Creation
- At Wikimania I presented my new Commons tool FastCCI, which is a fast category intersection tool with a simple interface. It's most accessible function is retrieval of Featured Pictures, Quality Images, and Valued Images from a Category and its subcategories below (down to arbitrary depth). The server backend for FastCCI is hosted on Wikimedia Labs and the source code is available on Github. As a direct consequence of my Wikimania presentation I put in a lot of effort into the tool in a pre-Wikimania development sprint. As my talk had a non-technical component, where I discussed the intricacies and oddities of the Commons category system I needed to gather statistics to substantiate my claims. These statistics include complex graph theoretical analysis of the Category graph, such a data on loops (whereas a category is contained by one of its subcategories). The analysis tools developed specifically for my talk will be used to generate up-to-date data on the state of the category graph. This data can guide maintenance tasks (taking the loops as a concrete example, many of the tight loops can be attributed to user errors).
- As a result of my talk I had several conversations with other Wikimedians which made excellent suggestions on further harnessing the FastCCI database to perform complex maintenance tasks. One such task is the cleanup of existing categories. On commons overcategorization, where an image is containe both in a category as well as subcategories of that category, is against policy. The FastCCI database could be of use to quickly detect and remedy such cases of overcategorization. I am currently working on a first implementation of a bot for this task.
- A second suggestion was a consistency check on the category tree itself. Composite categories such as Churches in Ravensburg should be contained (possibly by several levels of separation) in the higher level categories Churches and Ravensburg. This type of check (which requires a heuristic to deduce the high level categories) could be implemented very efficiently with the FastCCI database.
- Interactions with other users at Wikimania have also alerted me to a few bugs and missing features in the FastCCI Gadget (the frontend to the FastCCI database server). This resulted in a number of changes to implement these features/fixes.
- This was only the immediate impact of the conference attendance. A lot of the ideas that were generated at the conference still need some beeding time and tangible results will take more than just three weeks after the conference!
Connections
[edit]Good talk attendance (probably 50 people in the audience) plenty of questions asked. Got approached by several audience members after the talk with further questions, suggestions, and collaboration ideas. Maarten Dammers (User:Multichill) suggested a use for the FastCCI database backend to serve certain maintenance tasks on commons. This is described above.
At the hackathon I got a chance to hear about Vagrant and install it on my machine. I expect to look into the code of the Multimedia Viewer soon to figure out how to hook existing commons tools, such as the image annotator into it. About this I had conversations with people from the multimedia team, such as Gergo Tisza, Rob Lanphier, and Fabrice Florin.
Anything else
[edit]I uploaded my photos to commons and categorized them under commons:Category:Wikimania 2014 by User:Dschwen. My slides are uploaded to File:FastCCI_-_Taming_the_Commons_Category_Tree.pdf. Note that this is a preliminary conversion of my HTML slides into PDF.