W3C home > Mailing lists > Public > public-vocabs@w3.org > May 2014

Re: Better description for 'keywords' property

From: Stéphane Corlosquet <scorlosquet@gmail.com>
Date: Tue, 20 May 2014 11:54:09 -0400
Message-ID: <CAGR+nnGKVt58+LEOUvoDN_dkTtMhHdSw8_5uWCYkMUgj73Og_g@mail.gmail.com>
To: "martin.hepp@ebusiness-unibw.org" <martin.hepp@ebusiness-unibw.org>
Cc: Markus Lanthaler <markus.lanthaler@gmx.net>, Dan Scott <dan@coffeecode.net>, Dan Brickley <danbri@google.com>, W3C Web Schemas Task Force <public-vocabs@w3.org>
Good point, Martin. The patterns here to have a plural property to account
for multiple value is quite unusual in schema.org (happy to be proven wrong
here). A while back we renamed all the plural properties to singular
assuming that published would repeat that property as many time as needed
for value. That's a fine way of handling multiple values. 'keywords' is the
exception.

What do we recommend for people who want and can use the singular property
pattern to handle multiple keywords? I think we should leave 'keywords' for
that comma-separated string of keywords only, and recommend other
properties such as 'about' for individual keywords. Drupal 8 uses
http://schema.org/about whether it's a linked keyword or a plain test
keyword. We're not using 'keywords' since we prefer to preserve the data
structure, and it would actually require extra work to break the typical
one-property -> one-value pattern that we have generically implemented at
the moment. Of course if someone really wanted to use 'keywords', they
could always create a basic text field and map it to 'keywords' instead of
using Drupal's built-in taxonomy module...

Steph.


On Tue, May 20, 2014 at 11:37 AM, martin.hepp@ebusiness-unibw.org <
martin.hepp@ebusiness-unibw.org> wrote:

> Note that for an informed decision, we would have to look into the data
> structures driving typical dynamic Web sites. If they store single
> keywords, we can recommend single keywords per property. If they typically
> hold lists of keywords, we should recommend a string with a delimiter. If
> both is popular, allow both.
>
> It can be unnecessary burdensome for a Web developer to tokenize a string
> given from a back-end database in the template code with regular
> expressions or similar.
>
> So in a nutshell, a good Web vocabulary should support a dynamic degree of
> granularity - allowing site-owners to preserve as much structure as is
> available from the existing data sources, but not enforcing the lifting and
> cleansing of the data, since this will limit the amount of data published.
>
> Martin
>
> On 20 May 2014, at 16:45, Stéphane Corlosquet <scorlosquet@gmail.com>
> wrote:
>
> >
> >
> >
> > On Tue, May 20, 2014 at 10:39 AM, Markus Lanthaler <
> markus.lanthaler@gmx.net> wrote:
> > On Tuesday, May 20, 2014 4:00 PM, Dan Scott wrote:
> > > On Tue, May 20, 2014 at 02:17:12PM +0100, Dan Brickley wrote:
> > > >On 17 May 2014 06:31, Stéphane Corlosquet <scorlosquet@gmail.com>
> wrote:
> > > >> From previous conversations on this list, it looks like
> > > >> http://schema.org/keywords is meant to hold a list of
> comma-separated
> > > >> keywords, like the RDFa on this page:
> > > >> http://arc.lib.montana.edu/msu-photos/item/286:
> > > >>
> > > >> <span property="keywords">john burke, msc, football, team</span>
> > > >>
> > > >> If this is correct, the description for this property, which
> currently
> > reads
> > > >> "The keywords/tags used to describe this content", could be a bit
> more
> > > >> detailled. I suggest:
> > > >>
> > > >> A comma-separated list of keywords/tags used to describe this
> content.
> > > >
> > > >This sounds reasonable to me. The only objections I can think of
> > > >involve trying to stretch this property too far, e.g. phrases that
> > > >contain commas within them. Let's keep it simple...
> > > >
> > > >Does anyone here think that this change would not be an improvement?
> >
> > I was just wondering why there doesn't exist a singular version of
> > "keywords", i.e., "keyword". Was that somehow forgotten when all plurals
> > were deprecated or was this a deliberate decision?
> >
> > I had the same reaction as you at first when I discovered this, but
> 'keywords' was kept plural for that very reason, because it's one string
> containing a list of comma-separated keywords. I was surprised initially
> but apparently there are system/folks who prefer to use that as a opposed
> to breaking down the list into individual properties.
> >
> > Steph.
> >
> >
> > I think this matters because...
> >
> > > there are currently hundreds and, as sites upgrade, will be thousands
> > > of library Web sites that express "keywords" like:
> > >
> > > * keywords: Linux.
> > > * keywords: Internet programming.
> > > * keywords: Web sites > Design.
> > > * keywords: Electronic mail systems > Security measures.
> > >
> > > This is because we augment the existing display of subject headings
> like
> > > so:
> > >
> > > <div property="keywords">
> > >   <a href="search?email">Electronic mail systems</a> &gt;
> > >   <a href="search?email+security">Security measures.</a>
> > > </div>
> >
> > could also be expressed as
> >
> >   <span property="keyword"><a href="search?email">Electronic mail
> > systems</a></span> &gt;
> >   <span property="keyword"><a href="search?email+security">Security
> > measures</a></span>.
> >
> > which would have the advantage that the keywords are already tokenized by
> > the publisher instead of forcing the consumer to do so... which would,
> btw.,
> > also address Stéphane's concern below
> >
> > > >This sounds reasonable to me. The only objections I can think of
> > > >involve trying to stretch this property too far, e.g. phrases that
> > > >contain commas within them. Let's keep it simple...
> >
> >
> > --
> > Markus Lanthaler
> > @markuslanthaler
> >
> >
> >
> >
> > --
> > Steph.
>
>


-- 
Steph.
Received on Tuesday, 20 May 2014 15:54:37 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:29:41 UTC