Thanks, Charles. That's a very fine piece of work! And so relevant, not
just to the quiz use case but to the whole of the NLG side of our project.
I wonder whether there have been any further developments. I don't see any
links to that pdf on Google. I've put the link in our to-do list[1].
There's some good stuff in there about performance as well as data quality,
which are both areas we should certainly be looking into.
On the question of data structure, I guess it rather depends where you are
sitting. The essence of my quiZiverse idea is that the consumer handles a
relatively small dataset (client side) because WMF servers are running the
functions. Essentially, the result is a pretty manageable structured object
(in the more collaborative mode, perhaps a ZObject) and it can be grown
iteratively. Given those assumptions, it hardly matters which formats are
desired as inputs to (for consumption within) an actual quiz.
Since I was envisaging passing out "links" that are essentially a
pre-written call back to the function with different arguments, those
potential call-backs could be queued for processing server-side, so that
the eventual call-back is referencing a freshly minted structured object
(with a fairly limited shelf-life, unless it's a refresh of a pre-existing
ZObject). So, again, lots of options to explore on the technical side.
I'm inclined to disagree with you on the question of hints, though. The
structure I was envisaging is very straightforward; the "distractor" is an
answer to some other question, so the "hint" is just that question phrased
as a statement. It could be more complicated, but it should still be a
fairly simple connection to the next function call. Of course, "where
next?" may depend on whether the question was answered correctly, or there
might be a choice to be made, but I think that would still resolve quite
simply into a single next call. The functionality is an interactive
"crawler" [2], at the end of the day, with each "next step" deferred
until
required or pre-prepared if responsiveness might be an issue.
Keep it simple; iterate collaboratively; make it great!
Best regards,
Al.
[1]
https://meta.wikimedia.org/wiki/Talk:Abstract_Wikipedia/
Related_and_previous_work/Natural_language_generation
<https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Related_and_previous_work/Natural_language_generation>
[2]
https://en.wikipedia.org/wiki/Web_crawler
On Monday, 3 August 2020, <abstract-wikipedia-request(a)lists.wikimedia.org>
wrote:
Send Abstract-Wikipedia mailing list submissions to
abstract-wikipedia(a)lists.wikimedia.org
To subscribe or unsubscribe via the World Wide Web, visit
https://lists.wikimedia.org/mailman/listinfo/abstract-wikipedia
or, via email, send a message with subject or body 'help' to
abstract-wikipedia-request(a)lists.wikimedia.org
You can reach the person managing the list at
abstract-wikipedia-owner(a)lists.wikimedia.org
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Abstract-Wikipedia digest..."
Today's Topics:
1. Re: Comprehension questions (Charles Matthews)
2. Natural Language and Mathematics Generation (Adam Sobieski)
3. Re: Natural Language and Mathematics Generation (Charles Matthews)
4. Loose notes (Andy)
----------------------------------------------------------------------
Message: 1
Date: Sun, 2 Aug 2020 17:22:11 +0100 (BST)
From: Charles Matthews <charles.r.matthews(a)ntlworld.com>
To: "General public mailing list for the discussion of Abstract
Wikipedia (aka Wikilambda)" <abstract-wikipedia(a)lists.wiki
media.org>
Subject: Re: [Abstract-wikipedia] Comprehension questions
Message-ID: <29359200.234452.1596385332016(a)mail2.virginmedia.com>
Content-Type: text/plain; charset="utf-8"
On 02 August 2020 at 16:16 Grounder UK
<grounderuk(a)gmail.com> wrote:
But we don't just want the answer, we
want a quiz! Equally, maybe we
don't just want the question and the answer, we
want some wrong answers and
some tips.
It goes back to 2016, just to generate questions from Wikidata:
https://pub.tik.ee.ethz.ch/students/2016-FS/BA-2016-03.pdf
Technically the incorrect answers in multiple choice are called
"distractors". Clearly this is a rather simple data structure to handle.
Hints assume quite a bit more.
At the beginning of 2017, I decided to take seriously the suggestion (from
Magnus Manske) that questions should be treated as structured data. I even
suggested Wikidata should have a namespace for them (this didn't go down
well). A road not taken then, and just as the Comprende! tool was finished
I got diverted into a Wikimedian in Residence position. So much for that.
Anyone, one take on this is that AW output might be some kind of
structured data, rather than the sectioned prose (+media files and tables
and templated data) familiar from Wikipedia.
By the way, mathematics in wikitext has traditionally been a threefold mix
of approaches (HTML, png, LaTex): not an elegant solution.
Charles