[uf-discuss] Re: Microformats for Q&A sites

Thu Jan 21 19:06:50 PST 2010

Check out the definition [1] and principles [2] pages as well if you
haven't already. Pay close attention to what `microformats are not`.

[1]: http://microformats.org/wiki/microformats
[2]: http://microformats.org/wiki/principles

> Yes, I've been implementing hAtom as much as possible. In social Q&A sites (like Stack Overflow, which you mention), a question is analogous to a forum thread, which is episodic content, so that's how I'm structuring the microformats for Question2Answer. But I also need to indicate (a) - (c) ...

> (a) denote that a particular hentry is a question or an answer

Way too specific for hAtom considering an episode can be anything.
This is where it would help to see some HTML to better understand your
approach. Are/should the question and the answer(s) be contained in a
single `hfeed`, explicit or implicit? I'm thinking:

* home page
  * hfeed of popular questions
* question page
  * hfeed implicitly defined by a standalone hentry-wrapped question
  * hfeed exlicitly wrapping the answers
* user page
  * hfeed for questions
  * hfeed for answers

That's five feeds across three pages. Are we on the same page?

> (b) flag one particular answer as the one that was selected as the best

I highly doubt a microformat will ever contain a boolean `best`
property — especially not hAtom. "Best" == `highest rated` to me. How
about hReview [3]? Can you consider an answer's score to be a
`community review` of the answer? Check out hReview's `rating`
property.

> rating:: The rating is a fixed point integer (one decimal point of precision) from 1.0 to 5.0 inclusive indicating a rating for the item, higher indicating a better rating by default. Optionally a different integral "worst" value and/or "best" value MAY be specified to indicate a different range (e.g. 6 from 0-10). The "best" value may be numerically smaller than the "worst" value.

So each `hentry` of the `hfeed` defining a question page's answers
will include a community-generated `hreview` of itself. Assuming
you'll give each said `hentry` an `id` and permalink (rel-bookmark)
back to this via fragment identifier you should have no problem
pointing the `hreview.item` to the same fragment.

Now a machine that understands hAtom and hReview can grab the feed,
grab the reviews, pair the URIs, and infer the "best".

[3]: http://microformats.org/wiki/hreview

> (c) to show the net total of votes (any integer) received by a question or answer.

Check out hReview-aggregate [4].

[4]: http://microformats.org/wiki/hreview-aggregate

> Thanks - I'll look into that too. Another issue is that there are a bunch of important statistics on the user pages, that a Q&A search engine could use to build a reputation rating. Examples: number of questions/answers posted, number of votes given/received, total points. These are natural candidates for microformats, but I couldn't find any kind of prior work on this that I can draw on.

Number of Q/A posted requires just tallying hentries, no? If you need
to paginate them, provide `rel-me`s to the other pages and unify the
feeds via a singular `rel-tag` pointing to
http://answers.onstartups.com/users/123/alex-papadimoulis/answers
which will tag the feed as `answers` but will make sure that user
123's `answers` and user 498's `answers` are not confused to be of the
exact same feed category which might happen if you were to share the
tagspace of ..ups.com/answers.

`Total points` does /not/ strike me as a natural candidate for a
microformat. Be weary of making such claims. Definitely read about
`not boiling the ocean` and try to get us [me :)] some HTML to stare
at.

Welcome to the wonderful world of microformats.

Angelo

On Wed, Jan 20, 2010 at 10:45 PM, Gideon Greenspan
<research at gidgreen.com> wrote:
> Angelo,
>
> Thanks a lot for your reply. I thought I'd reply to the mailing list for
> now, since some of this might be of general interest.
>
>>> For example, a question page (that also shows related answers) will
>>> be a series of CLASS="hentry" elements. Some other elements that work
>>> well: entry-title, entry-content, rel=tag, published and author.
>>
>> To clarify, hAtom [1] specifies a standard for episodic content.
>> `hentry` is, in a sense, on par with `entry-title`, `entry-content`,
>> and the others. `hentry` alone is normatively meaningless without the
>> specified required properties - check out the cheatsheet [2] for a
>> quick reference if you haven't already.
>>
>> [1]: http://microformats.org/wiki/hatom
>> [2]: http://microformats.org/wiki/hatom-cheatsheet
>
> Yes, I've been implementing hAtom as much as possible. In social Q&A sites
> (like Stack Overflow, which you mention), a question is analogous to a forum
> thread, which is episodic content, so that's how I'm structuring the
> microformats for Question2Answer. But I also need to indicate (a) - (c)
> immediately below, which don't appear to have any place in hAtom.
>
>>> For example: (a) to denote that a particular hentry is a question
>>> or an answer, (b) to flag one particular answer as the one that was
>>> selected as the best, (c) to show the net total of votes (any ą
>>> integer) received by a question or answer.
>>
>> Check out the Q&A brainstorming [4], contribute in that dimly lit
>> corner, and elicit feedback here rather than poking at hAtom's spec.
>> Also consider using vote-links [5] via fragment identifiers. Confirm
>> the feasibility of that method via the spec. "Best" and "total"
>> shouldn't be too difficult to determine algorithmically once votes are
>> represented.
>>
>> [4]: http://microformats.org/wiki/question-answer-brainstorming
>> [5]: http://microformats.org/wiki/vote-links
>
> Thanks - I wasn't aware of those two pages, and I'll see what I can
> contribute to the Questions and Answers brainstorming page. Unfortunately
> vote-links doesn't look suitable, since I don't want to reveal *who* voted
> on questions and answers, just the aggregate information.
>
>> Lastly, mark up your home page with a representative hCard [6] and a
>> `rel-me` to your LinkedIn. Then implement the hAtom (and associated
>> hCard) semantics on said Q&A pages and do similar for the user pages,
>> validate everything, and don't hesitate to come back with questions or
>> feedback related to non-standard pursuits.
>>
>> [6]: http://microformats.org/wiki/representative-hcard
>
> Thanks - I'll look into that too. Another issue is that there are a bunch of
> important statistics on the user pages, that a Q&A search engine could use
> to build a reputation rating. Examples: number of questions/answers posted,
> number of votes given/received, total points. These are natural candidates
> for microformats, but I couldn't find any kind of prior work on this that I
> can draw on.
>
> Best regards,
>
> Gideon
>
>
> --
> Gideon Greenspan
> http://www.gidgreen.com/
> _______________________________________________
> microformats-discuss mailing list
> microformats-discuss at microformats.org
> http://microformats.org/mailman/listinfo/microformats-discuss
>

-- 
Angelo Gladding
angelo at gladding.name
http://angelo.gladding.name/