Re: "score" property for Q&A entities

On 16 June 2014 15:15, Alf Eaton <eaton.alf@gmail.com> wrote:

> On 16 June 2014 15:03, Dan Scott <dan@coffeecode.net> wrote:
>
>> On Mon, Jun 16, 2014 at 02:50:14PM +0100, Alf Eaton wrote:
>>
>>> On 16 June 2014 14:31, Dan Scott <dan@coffeecode.net> wrote:
>>>
>>>  On Mon, Jun 16, 2014 at 02:04:02PM +0100, Alf Eaton wrote:
>>>>
>>>>  On 16 June 2014 13:57, Dan Scott <dan@coffeecode.net> wrote:
>>>>>
>>>>>  On Mon, Jun 16, 2014 at 01:35:18PM +0100, Alf Eaton wrote:
>>>>>
>>>>>>
>>>>>>  In a discussion thread on this list a few months ago[1], there was
>>>>>>
>>>>>>> suggestion of adding a "score" property to Question/Answer/Comment
>>>>>>> (Q&A)
>>>>>>> classes, alongside the existing "upvoteCount" and "downvoteCount"
>>>>>>> properties.
>>>>>>>
>>>>>>> As I'm currently marking up Q&A pages that display only a score (and
>>>>>>> not
>>>>>>> counts of individual upvotes and downvotes) [2], this would be a
>>>>>>> useful
>>>>>>> property. Did the discussion ever turn into a full proposal?
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Alf
>>>>>>>
>>>>>>> [1] http://lists.w3.org/Archives/Public/public-vocabs/2014Feb/
>>>>>>> 0088.html
>>>>>>> [2] https://peerj.com/questions/31-what-does-open-access-mean-
>>>>>>> to-you/
>>>>>>>
>>>>>>>
>>>>>>>  In the example from peerj, the "score" that is being displayed has
>>>>>> nothing to do with upvotes or downvotes on the particular answer to a
>>>>>> question; it's the number of contributions that the individual
>>>>>> offering
>>>>>> that answer has made to the site as a whole (the sum of activity such
>>>>>> as
>>>>>> authored articles, edited articles, reviews, answers, questions, and
>>>>>> replies contributed).
>>>>>>
>>>>>> So I think "score" would be misleading if added to
>>>>>> Question/Answer/Comment for this particular example, because it is
>>>>>> attached to the person's account for that service. It seems more
>>>>>> appropriate for a social account property.
>>>>>>
>>>>>>
>>>>>
>>>>> The "scores" that I was referring to on the linked page are the numbers
>>>>> between the "thumbs up" and "thumbs down" icons, which are the scores
>>>>> that
>>>>> users have given to each question or answer, calculated as upvoteCount
>>>>> minus downvoteCount. This is the same as is displayed for questions and
>>>>> answers on StackOverflow, for example. I think you were probably
>>>>> looking
>>>>> at
>>>>> the numbers next to the authors of each question/answer, which are
>>>>> something else.
>>>>>
>>>>>
>>>> Oh, my apologies. Firefox blocks cross domain webfonts, so the
>>>> thumbs-up/thumbs-down appear as generic unicode boxes on that site and
>>>> obscured what I was supposed to be looking at :/
>>>>
>>>>
>>> My apologies for that - I'll try to get that fixed. Here's a better
>>> example, on StackOverflow:
>>> https://stackoverflow.com/questions/1960473/unique-values-in-an-array
>>>
>>>
>>>  Would there be a significant semantic difference for sites like this in
>>>> just treating the single score "X" as "X" upVotes, always with 0
>>>> downVotes?
>>>>
>>>
>>>
>>> It does feel like there's a difference, although the score could just be
>>> described as "net upvotes" in this case (in other cases, the score might
>>> be
>>> calculated differently).
>>>
>>> It partly depends on whether the data consumer is using "upvoteCount" as
>>> a
>>> measure of activity: an "upvoteCount" of "2" would imply little activity,
>>> when the item could actually have had 50 upvotes and 48 downvotes. It
>>> would
>>> also lead to the possibility of an item having "-10" upvotes, for
>>> example.
>>>
>>
>> Point taken.
>>
>>
>>  Is it worth overloading "upvoteCount" to the point where it loses its
>>> meaning, for the sake of not adding another property to the schema?
>>>
>>
>> My bias is towards conservatism and trying to explore the possibilities
>> of using the existing vocabulary before adding yet another property.
>>
>> The alternative seems to be to follow Martin Hepp's very sensible
>> suggestion of using the existing Review / Rating branch of the schema
>> vocabulary (with Rating's single ratingValue property) in this instance,
>> in which case you would have a very compact review / Review /
>> reviewRating / Rating / ratingValue nested set of properties and types
>> for the single score for each answer/comment/question. Pretty verbose I
>> guess.
>>
>> Optionally, one could simply expand the domain of ratingValue to include
>> Answer/Comment/Question, thereby still avoiding the addition of an
>> entirely new property, yet offering the score that you're looking for
>> along side the upvoteCount / downvoteCount properties.
>
>
> A "score" calculated from combined up/down votes does seem conceptually
> similar to a "quality" rating. It's more of an AggregateRating than a
> singular Rating, though (in which case "upvoteCount" and "downvoteCount"
> could be viewed as "positiveRatingCount" and "negativeRatingCount"). It
> differs from AggregateRating mostly in that the final score is a sum of all
> the votes rather than an average.
>
> I think it still seems different enough from either of the existing
> "rating" classes, as currently attached to reviews, that it would be a bit
> of a squeeze to re-use them.
>

In fact, I see now that the original Q&A schema research[1] suggested that
AggregateRating be used for the aggregate score (upvotes - downvotes).
AggregateRating is problematic when used as a sum of votes, though, because
there's no "bestRating" or "worstRating" and it's a total score rather than
an average rating.

[1] https://www.w3.org/wiki/WebSchemas/QASchemaResearch

Alf