Re: [Wikimedia-l] [Commons-l] Classification/Categorization games for Commons?

[Alex Stinson]

Two thoughts directly to this:

On Wed, Jun 6, 2018 at 3:43 PM, Gnangarra  wrote:

> I'd be more concern that the game throws up very generic and vague
> descriptions like, person, ship, cat, dog, tree, flowers, street. Which in
> itself might seem helpful but may not even highlight the important aspect.
>
Pushing images into narrower, but still generic, topic areas allows folks
who know more about, for example, ships or dog species or whatever, to
narrow the topic into something specific. We already do this in many ways
within the way folks use the category system on Commons -- it could be done
with structured data as well.

We would also want some type of confidence factor, I would think,
especially if we want the tool to appeal to newer folks in the community --
with less depth of experience working in our information structure.
Zooniverse and other similar visual-identification crowdsourcing projects,
usually have 2-3 volunteers confirm something before adding it directly to
the record.



>
>- I would hope many of the GLAMs have embedded keywords in the meta
>data/camera info which could be extracted like co-ords are.
>
> You would be surprised at how bad the metadata is in most collections --
only the most high profile collections will have good metadata (and one
perceived benefit of sharing GLAM content in public venues is the chance to
enrich metadata, by discovering details about the objects, that previously
the staff didn't know how to recognize). Wikimedia projects are really good
places to get mildly obscure collections (such as archival photos, or
under-researched museum objects) into the context of Wikimedia content.
Moreover, it's really hard to assess how to map these metadata concepts to
Wikimedia categories.  We actually saw this come up again and again in the
GLAM stakeholder research for commons:
https://meta.wikimedia.org/wiki/Research:Supporting_Commons_contribution_by_GLAM_institutions



>
> On 7 June 2018 at 02:07, Alex Stinson via Commons-l <
> common...@lists.wikimedia.org> wrote:
>
>> Hey Yaroslav and Asaf,
>>
>> From the SDC team perspective, I think it would definitely be preferable
>> for such a tool to include/anticipate the need for structured data on
>> Commons, or default to filling in Depicts and/or other structure data
>> fields. Building more tools which generate categories by default would
>> definitely be a bit counter-productive (and hard on multilingual
>> contributors). One option, might be designing such a tool to work with
>> Artworks and other unique objects (like photographs) already on Wikidata,
>> and then have it prepared to hook up with the  Wikibase/Structured data
>> features that will go live on Commons in the fall.
>>
>> Cheers,
>>
>> Alex
>>
>>
>>
>> On Wed, Jun 6, 2018 at 2:01 PM, Asaf Bartov 
>> wrote:
>>
>>> Absolutely; I wanted to know if such a tool perhaps already exists.  If
>>> one
>>> does not, then definitely, if we develop a tool, it should look to the
>>> future and be based on Structured Data on Commons already!
>>>
>>>A.
>>>
>>> On Wed, Jun 6, 2018 at 8:54 PM Yaroslav Blanter 
>>> wrote:
>>>
>>> > I think it is pretty similar to what we have built in Wikidata, Do
>>> > Structured Commons folks want to comment?
>>> >
>>> > Cheers
>>> > Yaroslav
>>> >
>>> > On Wed, Jun 6, 2018 at 7:47 PM, Asaf Bartov 
>>> wrote:
>>> >
>>> > > Hi, folks.
>>> > >
>>> > > It occurs to me there are tens or hundreds of thousands of images
>>> donated
>>> > > en masse (GLAM etc.) that are only categorized as "image from X
>>> > collection"
>>> > > or "Files donated by X", i.e. essentially uncategorized by content.
>>> > >
>>> > > This obviously greatly reduces the likelihood of discoverability and
>>> > > re-use.  But it's hard to find such files, and the massive categories
>>> > > (thousands of files, often) don't make organizing the work easy.
>>> > >
>>> > > I'm think of a gamified interface -- à la Wikidata Game -- that would
>>> > let a
>>> > > volunteer (after OAuth identification) pick a category (from a
>>> pre-fed
>>> > list
>>> > > of massive categories of donated files) and show one photo from the
>>> > > category that has only that category listed (i.e. has no
>>> categorization
>>> > by
>>> > > content), and let the volunteer type (with auto-complete, like
>>> HotCat)
>>> > some
>>> > > appropriate categories and hit Save, and the categories would be
>>> added,
>>> > and
>>> > > the next file shown.
>>> > >
>>> > > (Optionally, a second layer of verification could be added, where
>>> > > volunteers would [also] be invited to vet or change previous
>>> volunteers'
>>> > > categorization, and actual change to categories on Commons would only
>>> > take
>>> > > place after 2 (or N) users approved the categories.  I'm not at all
>>> sure
>>> > > this is needed, and I think we can start without it and see how it
>>> goes.)
>>> > >
>>> > > So, does something like this exist?  If not, who wants to build it?
>>> :)
>>> > >
>>> > >A.
>>> > > 

Re: [Wikimedia-l] [Commons-l] Classification/Categorization games for Commons?

[Gnangarra]

I'd be more concern that the game throws up very generic and vague
descriptions like, person, ship, cat, dog, tree, flowers, street. Which in
itself might seem helpful but may not even highlight the important aspect.

   - Would a bot doing an initial keyword search of the description be more
   effective
   - descriptions should already have a language code embedded.
   - I would hope many of the GLAMs have embedded keywords in the meta
   data/camera info which could be extracted like co-ords are.


On 7 June 2018 at 02:07, Alex Stinson via Commons-l <
common...@lists.wikimedia.org> wrote:

> Hey Yaroslav and Asaf,
>
> From the SDC team perspective, I think it would definitely be preferable
> for such a tool to include/anticipate the need for structured data on
> Commons, or default to filling in Depicts and/or other structure data
> fields. Building more tools which generate categories by default would
> definitely be a bit counter-productive (and hard on multilingual
> contributors). One option, might be designing such a tool to work with
> Artworks and other unique objects (like photographs) already on Wikidata,
> and then have it prepared to hook up with the  Wikibase/Structured data
> features that will go live on Commons in the fall.
>
> Cheers,
>
> Alex
>
>
>
> On Wed, Jun 6, 2018 at 2:01 PM, Asaf Bartov  wrote:
>
>> Absolutely; I wanted to know if such a tool perhaps already exists.  If
>> one
>> does not, then definitely, if we develop a tool, it should look to the
>> future and be based on Structured Data on Commons already!
>>
>>A.
>>
>> On Wed, Jun 6, 2018 at 8:54 PM Yaroslav Blanter  wrote:
>>
>> > I think it is pretty similar to what we have built in Wikidata, Do
>> > Structured Commons folks want to comment?
>> >
>> > Cheers
>> > Yaroslav
>> >
>> > On Wed, Jun 6, 2018 at 7:47 PM, Asaf Bartov 
>> wrote:
>> >
>> > > Hi, folks.
>> > >
>> > > It occurs to me there are tens or hundreds of thousands of images
>> donated
>> > > en masse (GLAM etc.) that are only categorized as "image from X
>> > collection"
>> > > or "Files donated by X", i.e. essentially uncategorized by content.
>> > >
>> > > This obviously greatly reduces the likelihood of discoverability and
>> > > re-use.  But it's hard to find such files, and the massive categories
>> > > (thousands of files, often) don't make organizing the work easy.
>> > >
>> > > I'm think of a gamified interface -- à la Wikidata Game -- that would
>> > let a
>> > > volunteer (after OAuth identification) pick a category (from a pre-fed
>> > list
>> > > of massive categories of donated files) and show one photo from the
>> > > category that has only that category listed (i.e. has no
>> categorization
>> > by
>> > > content), and let the volunteer type (with auto-complete, like HotCat)
>> > some
>> > > appropriate categories and hit Save, and the categories would be
>> added,
>> > and
>> > > the next file shown.
>> > >
>> > > (Optionally, a second layer of verification could be added, where
>> > > volunteers would [also] be invited to vet or change previous
>> volunteers'
>> > > categorization, and actual change to categories on Commons would only
>> > take
>> > > place after 2 (or N) users approved the categories.  I'm not at all
>> sure
>> > > this is needed, and I think we can start without it and see how it
>> goes.)
>> > >
>> > > So, does something like this exist?  If not, who wants to build it? :)
>> > >
>> > >A.
>> > > ___
>> > > Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/
>> > > wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/
>> > > wiki/Wikimedia-l
>> > > New messages to: Wikimedia-l@lists.wikimedia.org
>> > > Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l
>> ,
>> > > 
>> > ___
>> > Wikimedia-l mailing list, guidelines at:
>> > https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and
>> > https://meta.wikimedia.org/wiki/Wikimedia-l
>> > New messages to: Wikimedia-l@lists.wikimedia.org
>> > Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
>> > 
>> ___
>> Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wik
>> i/Mailing_lists/Guidelines and https://meta.wikimedia.org/wik
>> i/Wikimedia-l
>> New messages to: Wikimedia-l@lists.wikimedia.org
>> Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l,
>> 
>>
>
>
>
> --
> Alex Stinson
> GLAM-Wiki Strategist
> Wikimedia Foundation
> Twitter:@glamwiki/@sadads
>
> Learn more about how the communities behind Wikipedia, Wikidata and other
> Wikimedia projects partner with cultural heritage organizations:
> https://outreach.wikimedia.org/wiki/GLAM
>
>