The on-wiki version of this newsletter can be found here:
https://meta.wikimedia.org/wiki/Abstract_Wikipedia/Updates/2023-08-31
Wikifunctions at Wikimania 2023
We presented Wikifunctions at *Wikimania 2023*
<https://wikimania.wikimedia.org/wiki/2023:Wikimania>, the annual gathering
of Wikimedians from all around the world. This year, Wikimania saw
participants from 138 languages. The conference was held from 16 to 19
August in Singapore.
<https://meta.wikimedia.org/wiki/File:Wikimania_2023_Singapore_Blue_Header_logo.svg>Wikimania
2023 in Singapore, August 16-19
It was exciting to see the Wikifunctions logo featured during Wikimania as
part of logos of the Wikimedia projects on the materials and big screens -
including on the largest high definition video wall in the world, according
to the Guiness Book of Records. It was even more exciting to present
Wikifunctions to the Wikimedia community.
We had two sessions, one that featured a presentation and question and
answer session, *Wikifunctions is here*, where we discussed where we are
and our plans, and were listening to the ideas and requests from the other
projects. A recording of this session is available, split in two parts with
a slight gap due to technical difficulties (one
<https://www.youtube.com/watch?v=sKoMYHnZhEk&t=20046s>, two
<https://www.youtube.com/watch?v=ldHYoD_1j5I>). The Wikimania organizers
are still working on uploading the individual parts to Commons, and we hope
to have an unbroken recording in the future.
The second session, *First steps with Wikifunctions*, was a tutorial with
the goal of taking contributors to work on their first function. We are
very happy about how the tutorial went. All participants managed to create
their first function on Wikifunctions. We are particularly happy about the
people who self-described themselves as non-developers also being able to
successfully follow the tutorial and work on functions, tests, and
implementations themselves. We certainly identified steps that need
improvement, particularly for creating tests and for the selection of
arguments in implementations, and we are grateful for these having clear
ways forward. A recording of the first part
<https://www.youtube.com/watch?v=YiwjqxqSENs&t=24940s> of the session is
available, but we did not record the working part of the session.
We want to thank the organizers of Wikimania for providing us with a stage
to present Wikifunctions, and the attendees who gave such valuable feedback!
CoSMo language
Members of Abstract Wikipedia’s natural language generation special
interest group (NLG SIG) under the lead of Maria Keet and Kutz Arrieta have
worked on the CoSMo content selection modeling language. CoSMo is a
language that enables the specification of which pieces of structured data
to use to drive the generation of a piece of text. This allows to decide
what part of the billions of statements in Wikidata would be considered
relevant for the generation of a given piece of text in Abstract Wikipedia.
Maria describes CoSMo and its background further in her blog post “CoSMo: a
content selection modelling language
<https://keet.wordpress.com/2023/08/20/cosmo-a-content-selection-modelling-language/>”
on the Keet blog. The full tech report is available on Arxiv
<https://arxiv.org/abs/2308.02539>.
GFpedia demo
Over the last few weeks, Krasimir Angelov, one of the developers of
Grammatical Framework (GF), and member of the NLG SIG, has developed
GFpedia, a system to demonstrate how parts of Abstract Wikipedia could work
using GF technology.
A few example pages:
- The Netherlands in English
<https://cloud.grammaticalframework.org/wikidata/index.wsgi?id=Q55&lang=en>
- Sweden in Bulgarian
<https://cloud.grammaticalframework.org/wikidata/index.wsgi?id=Q34&lang=bg>
Grammatical Framework <https://en.wikipedia.org/wiki/Grammatical_Framework> is
a programming language focused specifically on writing grammars for natural
languages, both for parsing and generation. In this case, the demo calls
the data from Wikidata and uses it to drive the generation of text. It is
using a template per type of entity, i.e. you will find the same template
used for every country, whereas many other types have smaller or no
templates currently.
CCKS keynote on Wikidata and Wikifunctions
On Sunday, Denny gave the keynote presentation to the 17th China Conference
on Knowledge Graphs and Semantic Computing <https://sigkg.cn/ccks2023/en/> in
Shenyang, China. In the keynote he presented Wikidata and Wikidata’s
lexical extension as well as Wikifunctions. The pre-recorded presentation
is available on YouTube <https://www.youtube.com/watch?v=MXEctjXD_GU>. The
questions in the Q&A session focused entirely on Wikifunctions, and the
interest in our new project was delightful.