[Gendergap] Commons Searches

Fri Oct 14 00:47:12 UTC 2011

Unfortunately we currently have zero developers working on search (as 
far as I know). There are several more significant search bugs that are 
also not going to be fixed any time soon. Another issue is that our 
search engine is Java while the rest of MediaWiki is PHP. This makes 
sense for performance reasons, but makes the pool of potential 
developers who are able and willing to work on it much smaller. In other 
words, this might get fixed in a few years, but I wouldn't hold my 
breathe. In the meantime, it would be good to follow Sarah's lead and 
proactively curate the content we have so that there is less potential 
for astonishment in our search results.

Ryan Kaldari

On 10/13/11 5:37 PM, Andreas Kolbe wrote:
> John,
>
>     *From:* John Vandenberg <jayvdb at gmail.com>
>     > (Searching for "levee" in Commons brings up an image of a
>     > naked Suicide Girl called Levee in third place.)
>
>     Its a thumbnail for !@#$ sake, and anyone who finds that image
>     offensive should turn off their internet connection.
>
>
> It's a perfectly nice image, but does it answer the user's need? In 
> most cases probably not. If I google levee, I see levees, not nude girls:
>
> http://www.google.co.uk/search?gcx=c&q=levee&um=1&ie=UTF-8&hl=en&tbm=isch&source=og&sa=N&tab=wi&biw=1041&bih=638 
> <http://www.google.co.uk/search?gcx=c&q=levee&um=1&ie=UTF-8&hl=en&tbm=isch&source=og&sa=N&tab=wi&biw=1041&bih=638>
>
> If I want to google for pictures of Levee, I google for "Levee Suicide 
> Girls", and there she is:
>
> http://www.google.co.uk/search?gcx=c&q=levee&um=1&ie=UTF-8&hl=en&tbm=isch&source=og&sa=N&tab=wi&biw=1041&bih=638#um=1&hl=en&tbm=isch&sa=1&q=levee+suicide+girl&pbx=1&oq=levee+suicide+girl&aq=f&aqi=&aql=&gs_sm=e&gs_upl=127182l129981l0l130379l15l15l0l11l0l0l291l930l0.1.3l4l0&bav=on.2,or.r_gc.r_pw.,cf.osb&fp=120e52a58330422e&biw=1041&bih=638 
> <http://www.google.co.uk/search?gcx=c&q=levee&um=1&ie=UTF-8&hl=en&tbm=isch&source=og&sa=N&tab=wi&biw=1041&bih=638#um=1&hl=en&tbm=isch&sa=1&q=levee+suicide+girl&pbx=1&oq=levee+suicide+girl&aq=f&aqi=&aql=&gs_sm=e&gs_upl=127182l129981l0l130379l15l15l0l11l0l0l291l930l0.1.3l4l0&bav=on.2,or.r_gc.r_pw.,cf.osb&fp=120e52a58330422e&biw=1041&bih=638>
>
> I guess Commons should give more weight to categories, and less weight 
> to file names. So when I google cucumber, it should show me images in 
> the cucumber category first of all, and not images that happen to have 
> cucumber in the title.
>
> Brandon, is there something developers could do in this regard?
>
>
>     I am sure you'll be appalled that libraries include nude pictures in
>     their search results, often when searching for something else.
>
>     http://trove.nla.gov.au/picture/result?q=contemporary+north+america+20th+century
>
>     fix the metadata.
>
>     create a gallery page.
>
>     create a category and populate it.
>
>     etc
>
>     p.s. abstract art offends me.  Can we please remove media related to
>     John Levee's from the Commons search results for the term 'Levee'. ;-)
>
>     > We should be under no illusion that we can find all search terms
>     whose
>     > results violate the principle of least surprise, presenting adult
>     images for
>     > everyday search terms.
>     >
>     > New such situations arise on a daily basis, each time someone
>     uploads an
>     > explicit file that has a plausible search term in its name and
>     > description (try searching Commons for "eating", and then search for
>     > "drinking"; or try finding images of Prince Albert).
>
>     The ordering of the search results isnt ideal.  Have you raised a bug?
>
>
>
> The thing is, John, it's not a bug. How is it a bug? The image is 
> called "Drinking urine" or whatever, and so it's a valid search result 
> for "drinking". No doubt, a bunch of people would argue that it would 
> be non-neutral to exclude it from the search results for drinking, 
> because Wikipedia is not censored, and we don't care if people are 
> unhappy with our service, because that would be non-neutral. ;)
>
> <Imagine rant here.>
>
>
>     It puts too much weight on the filename, which isnt good because
>     recommend against rename, so the current search results are gamable by
>     the uploader.
>
>     > We should simply offer safe search, like Google does.
>
>     Google provides safe search.  They need to convert 'the internet' into
>     a search results page that their customer wants to see, and the
>     Internet has a whole lot of stuff that 99% of the world never wants to
>     see.
>
>     Wikipedia provides encyclopedic information.
>
>     Commons provides a depository of media, and if you search for keywords
>     in the metadata you'll see thumbnails of the matching media.
>
>
>
> I find Google safe search seriously useful, because it gives me a 
> choice, and enables me to tailor my search to my requirements. If I 
> want to see porn, I can see porn. If I'm looking for something else, I 
> can prevent my search being flooded with porn.
>
> If I am a researcher looking for images of Prince Albert on Commons, I 
> would appreciate not being forced to wade through dozens of images of 
> penises with rings in them to find the image I'm looking for.
>
> http://commons.wikimedia.org/w/index.php?title=Special:Search&redirs=1&ns0=1&ns6=1&ns9=1&ns12=1&ns14=1&ns100=1&ns106=1&search=Prince+albert&limit=500&offset=0 
> <http://commons.wikimedia.org/w/index.php?title=Special:Search&redirs=1&ns0=1&ns6=1&ns9=1&ns12=1&ns14=1&ns100=1&ns106=1&search=Prince+albert&limit=500&offset=0>
>
> We will not attract a more mature audience until we get our act together.
>
> Andreas
>
>
>
> _______________________________________________
> Gendergap mailing list
> Gendergap at lists.wikimedia.org
> https://lists.wikimedia.org/mailman/listinfo/gendergap
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.wikimedia.org/pipermail/gendergap/attachments/20111013/e8617488/attachment-0001.htm 