Wikipedia:GLAM/Museum of New Zealand Te Papa Tongarewa/What We've Done/Wikimania 2023

Post-conference summary from Avocadobabygirl and Stitchbird.

The Whole GLAM package
Our presentation on a workflow for GLAMs to share packages of open collection images, data, and knowledge was well attended and received, with several people following up with us for more conversation.





We're hopeful that more people will get in touch on Wiki or email – the important thing is to keep the work active and visible so it’s available when they need it.

In the week following the presentation, the manual’s landing page was viewed almost 1300 times.

Highlights
As well as the parts of the conference that provided concrete benefits and will lead to future engagement, Wikimania provided a whole suite of great experiences. Here’s just a few.

Other kiwis
A large contingent of New Zealanders attended Wikimania this year. We were deeply impressed by what the other kiwis brought – fascinating presentations, great conversations, and overall presence. No doubt many people from around the world met not just their first but second and third New Zealander and came away with new ideas.

Tala Wiki (WikiProject page) supported by Pakoire gave a phenomenal performance that combined music, dance, video, and genuinely biting satire.

DrThneed talked about the, presenting clear guidance on how to not just share information on Wikidata but also what you can then do with it. Jetaynz showed to support the Aotearoa New Zealand Histories Curriculum and hinted at what’s coming up.

Ambrosia10 continued to be everywhere getting into discussions with everyone to advance projects and solve problems, and was recognised for her contributions to the overall world of Wiki when she was announced as Wikimedian Laureate of the year!

Giantflightlessbirds provided perspective on being a Wikipedian in Residence in Aotearoa at the WREN panel. Chocmilk03 was part of a panel on Women in Red, where she’s an active contributor, particularly with women in New Zealand literature.

As a group we provided great mutual support, enthusiastic applause, and helpful introductions to contacts. Collaborative notes from the conference are on Wiki.



Community archive


MyCommunity is a very ground-up community heritage effort based in the Queenstown area of Singapore, which contrasts itself with standard (particularly in the region) top-down work that has led to public disengagement with heritage and archives.

Built on the back of a sustained door-knocking campaign, the project has a collection, museum space, on-site tours, hundreds of volunteers, a core of heritage professionals, and many direct links with residents of Queenstown. Collections come from residents after sustained relationship-building, and residents also have a steer in cataloguing and interpretation.

The project is working on using Wikibase as a collections management system, which will have a public front end. The site will draw on Wikidata to help describe and link collection objects, with the ability to build further connections and narratives. At the moment media of collections looks like it’ll still be in overall Wikimedia Commons.

Wikibase is being supplied as a cloud service by Wikibase.de.

Using AI to write Wikipedia articles
Artificial Intelligence (AI) including chat bots, ChatGPT, and other large language models (LLMs) were a common topic in many of the sessions at Wikimania. Many talks mentioned the tendency of such AI to “hallucinate” and that the results and citations were often fabricated, such that it cannot be trusted or used, for example, for writing Wikipedia articles or Wikidata items.

In that context it was interesting to go to the talk called, “Empowering Wikipedia and Wikidata with AI: Scholarly publications”. The author agreed that we should not simply allow LLMs to create a Wikipedia page from scratch, without any limitations. Instead, an editor should do a search about the topic to create a relevant bibliography, and then get the LLM to make main bullet points about each reference and only use these references in the bibliography for the article. The LLM should then be instructed to generate a Wikipedia article from all of these bullet points (and only these), and reference every statement.

In Wikidata, the LLM should provide the Wikidata ID corresponding to every considered concept and property, as well as the URLs of all the references, to avoid fabrication.

In this way, a good first draft of an article in Wikipedia (or an item in Wikidata) can be written with the help of AI, which of course then needs to be edited by a human.

GLAMS on Wiki and analytics
Though it was disappointing to realise that there isn’t a comprehensive toolkit for measuring the impact of a GLAM’s contributions, it was useful to understand that this is an issue affecting all institutions trying to work in Wiki, and we’re all struggling in the same ways.

There was a lot of discussion about the breakdown of available analytics tools and the unsuitability of the overall tool landscape for GLAMs, which is now having a concrete impact as major organisations including the Metropolitan Museum of Art are pulling their funding for actively sharing on/getting involved with Wiki. Smithsonian appears to be the only major US organisation with dedicated funding remaining.

There is no consolidated plan or support for GLAM engagement with Wiki from Wikimedia Foundation. GLAMs needs aren’t well understood/articulated (particularly if trying to think about using Wiki outside of a social media perspective) and while there’s a developing GLAM dashboard it’s not well-known or clear if it is being actively developed.

Auckland Museum uses GLAM dashboard, and we’ve requested access to it for Te Papa as well.

There will be a GLAM Wiki conference in Uruguay in November 2023, so we’ll be watching for what comes of that.

Mātauranga Māori and other local knowledge
We contributed to the many threads of discussion about how Indigenous knowledge(s) can interface with/be supported by Wiki. Although Wiki is founded on the idea of ‘universal access to all the world’s knowledge’, the New Zealand participants added the local perspective that mātauranga Māori isn’t material that should automatically be added to Wiki by Pākehā/institutions. This can be an issue given Wiki requires open licensing, and copyright remains the only real control on access and use.

Our impression is that for most participants the founding concept of universal access is still a given, and considering the complexities of knowledge where universal access (and thereby use/reuse) isn’t the wanted outcome will continue to require some gear-changing whenever it comes up. However, it seems to get some traction when discussing a group’s autonomy to choose what to share. This could be a lens to share information about WAI262, for example, with an international audience.

The performance by Tala Wiki did a brilliant job of raising similar points in a Pasifika context, pointing directly at the extraction of Indigenous knowledge from the Pacific to feed into a colonial machine. We also got fantastic examples from members of other minority or underrepresented groups who were directly using Wiki to both support their own communities and share with the wider world, for example: Aromanian, Romani, Macedonian sign language. Challenges include lack of written materials and lack of standardise written form for some languages. It’s super important to survey the community and community organisers to fully understand their goals and ideas prior to starting a new project, that may involve workshops, editing contests, seminars and recording sessions.

In some of the languages mentioned above, the focus was on creating content for their own communities (e.g. Macedonian sign language) whereas other communities were more focused on sharing their knowledge on English-language platforms because their goal is to increase knowledge and minimise misinformation with the wider Wiki audience (e.g. Romani).

Gender
Gender and other gaps in Wikiprojects were a major theme of the conference. On Tuesday 15 Aug 2023, Stitchbird2 was selected to attend the “Mind the Gap” full day session at Google Asia Headquarters. Google and Wiki projects point to each other frequently and have similar vision statements.

The latest statistics show that only one-fifth of Wikipedia biographies are about women (and less than one-fifth of those have a photo), less than one-quarter of active Wiki editors are women (up from one-tenth in 2011), about one-third of readers of Wikipedia are women, and for every one photo of a woman on Commons, there are 3 photos of men.

At Mind the Gap, an animated panel discussed “Empowering Women in Wikimedia” with a call to action to build courage, not confidence; collaborate with like-minded people to support your local community; take small steps to make big ripples; and lift yourself and others. Other sessions later at the conference to understand, discuss and remedy the gender gap (as well as celebrate women and other minorities in various Wikiprojects) included Visible Wikiwomen, Wikiwomen lunch, Women in Red, among others.

It’s worth noting that if we want to be helpful here, we’ve got back-end work to do as well. For example, of the person party records available (and with data in the gender field) on Te Papa’s Collections Online, 15,335 (70%) are male, 6,645 (30%) are female, and 15 are gender diverse. 17,812 records have no gender recorded.

Release of corporate images
Various presentations and conversations have prompted us to look again at a process for releasing corporate (non-collection) images for use in Wiki. We’ll shortly be doing a pilot release to help us figure out a repeatable workflow. At the Visible WikiWomen session, we discovered that it's important to not only upload images of women into Wikimedia Commons, but to make sure the images have structured data so they can be found. Some people may be at risk by being visible on Commons, so consent is essential, and there are guidelines and trust and safety tools available from Visible WikiWomen.

To help overcome the gender gap, we at Te Papa can implement several of the solutions proposed in an excellent article published in the journal Nature in 2022, "How academic institutions can help to close Wikipedia’s gender gap”:
 * Elevate and broadly feature the achievements and lives of women and people from gender minorities in Emu, Collections Online, bogs, narratives, hubs, the Te Papa website, press releases, etc. to help meet Wikipedia’s notability criteria for scientists and academics.
 * Upload more images to Wikimedia Commons of women scientists, collectors, curators, collection managers, artists, technicians, and makers (who work at Te Papa and elsewhere), with appropriate licences, descriptions and structured metadata to increase discoverability.
 * Host and support targeted edit-a-thons and build a community of support, especially to increase biographies about women, members of the LGBTQ+ community and other underrepresented groups.
 * Build an institutional Wiki-editing community in which contributing to Wikipedia is part of the broader institution organisation, to collectively chip away at Wikipedia’s gender, racial, geographical and societal bases.

Supporting communities
Sessions mentioned above have provided impetus to see how Te Papa can support communities more directly, and ideas about what that can look like. Several of our projects so far have been broadcast in nature – we find collections, data, and/or research that is useful to share and get it out there. That’s useful, but we should also be able to work with members of communities who have their own goals and priorities in mind, and find ways to support them.

We’re getting started on arranging a short set of edit-a-thons with community groups from our priority audiences, figuring out what Te Papa has (and has access to) that can support their goal of improving access to and the representation of their members and cultures. The events and the shape of the project itself are being influenced by what we learned at the conference.

Other staff at Te Papa are also getting interested in how Wiki might be useful for the communities they work with. More Te Papa projects are being explicitly community-led, and curators want to know if Wiki could be part of the toolkit we offer.

Wikisource
We had the chance to try Wikisource for the first time, which may be a viable place to share archival full-text objects. Te Papa holds a lot of this sort of material, including already-digitised material like our annual reports dating from the founding of the Colonial Museum.

Materials (text, pdf, djvu) can be loaded into Wikisource via Wikimedia Commons, at which point they are OCR’d and the community can view the text and make and validate corrections. When completed texts can be downloaded.

The OCR system is just getting an upgrade for recognising handwriting by implementing models from Transkribus, a machine learning system that operates across languages and character sets. By providing image files and accurate transcripts covering around 5000 words (a video tutorial is available), the resulting model can be used on a body of similar material and has error rates between 0.5-6% depending on the text. It’s most accurate when dealing with a single writer all in the same handwriting. Wikisource Loves Manuscripts can provide support.

As a machine learning (ML) system, the source files need to be uploaded into the platform, and rights/data retention need to be looked into. Our best bet if we use this will be to use English-language texts, especially archival material.

Onboarding institutions
Wiki Italia has been piloting an effort to get basically every GLAM institution onto Wiki by systematically making contact and providing a framework for providing at least 10 CC0 images - collections, interpretation panels, building photos. They have lots of structural stuff in place including a contact/getting started form they'll be translating and some neat mapped analytics.

They've improved Wikidata items for/related to thousands of orgs, 91 started the process, 22 have shared images, and they've worked up 5 case studies.

Avocadobabygirl spent a while talking to Iolanda Pensa from the project as Aotearoa already has a crucial part of this in place – the directory on Kōtuia. We could look at working with their team to test the system in another language/place, especially since NZ is nicely contained. It may even be better to focus on information about the organisations rather than their collections as it might be easier to apply open licenses.

We're now comparing Kōtuia's data with what's in Wikidata to see if we can make the sort of bulk updates that would provide a baseline for further work.

User/editor support
New user onboarding and retention appears to be a perennial topic. A session on “Mentoring new editors on Wikipedia” was however somewhat disappointing because we thought it would be about in person mentoring whereas it was about the Growth membership tools (including the “Mentor Dashboard”) available to do online mentoring on Wikipedia.

Anyone who signs up to be a mentor will get a Mentor Dashboard to help and encourage newcomers, and stay in touch with them. By providing online positive reinforcement, it is hoped this will lead to editor retention and productivity. However, the statistics did not show that online mentoring is actually achieving these goals in English or Spanish Wikipedia.

Our experience is that in person mentoring is extremely valuable. We are both extremely grateful to the mentoring that has been given to us by several Wikimedians in Aotearoa, but especially Ambrosia10 (our very own 2023 Wikimedia Laureate!), Giantflightlessbirds, and Einebillion. Their constant support, specific guidance and continual constructive feedback has been key to us entering the Wiki ecosystem and growing there.

Avocadobabygirl and Stitchbird2 have been mentoring another Te Papa natural history curator, Porrhothele, in his first Wikipedia articles about arachnids, and we are keen to continue encouraging and mentoring others in this space. We continue to hold in person fortnightly informal wiki lunch editing sessions at Te Papa for any staff who are interested and are hoping we can retain and mentor some additional editors from our staff.

Open access


The session on Open Climate Campaign was interesting, where we learned that open sharing of scientific research outputs is not the default as only about 40% of scientific papers are open access (OA) in the USA & Canada, and about 47% of climate change research articles are OA.

We learned that even if your published paper is not OA, there is a legal way to upload the accepted version of your manuscript to an online repository for self-archiving.

Observations
Stitchbird2 was one of seven Wikimanians participating in the Wikimania 2023 Bioblitz on iNaturalist whilst in Singapore, which has over 1,000 observations of some 400 species! She’s since joined the Wikiproject Biodiversity channel on Telegram and is inspired by the people in that international group and what they are doing and sharing in this space.

There was some talk on that Telegram channel about assessing all of the iNaturalist photos taken during the Bioblitz regarding whether those species had photos on Commons, and if not, bulk uploading them. Stitchbird2 is very interested in learning how to do this not only for these Bioblitz photos, but also for other similar datasets. Although we've used the iNaturalist2Commons tool for individual species, it would be good to know if there is a pipeline for doing multiple species at once for a particular iNaturalist dataset.

Stitchbird2 is newly inspired to keep doing species pages on Wikipedia (as well as updating Wikidata with relevant scientific references and scientists). Now that she has finished 65 species pages for all southern hemisphere Myosotis species (completed just in time for Wikimania 2023), she is planning on continuing this work on the other genera she has expertise in, namely Plantago, Ourisia and cushion Veronica. She also plans to trial and use some of the translation tools highlighted at Wikimania 2023 to help translate any pages she creates of relevant South American species of these three genera into Spanish.

And the rest!
Other engagement ideas we heard about at Wikimania that we may trial at Te Papa to engage Te Papa staff in Wikimedia events include:


 * #1lib1ref: This is not new to us, but last year’s full day event at Te Papa for Te Papa staff was highly successful and well worth repeating again.


 * #1pic1art: Along a similar vein to #1lib1ref, this project may also be another successful way to engage with staff who can add Commons photos to Wikipedia articles.


 * WikiStories is new Wikimedia Project that may be a good way to get new editors engaged as it is quite short form writing, not as involved/daunting as a full article.


 * I Am Remarkable is a 90-minute Google course focuses on the importance of self-promotion in your personal and professional life, equipping the student with tools to develop this skill and challenging the social perception around self-promotion. We would like to first take the course ourselves and then, if relevant, investigate how we might encourage other Te Papa Wikimedians and staff to also take part.


 * The Diff blog has been around for some time, but it was a good reminder to regularly check out the highlights here as to what is new in the Wiki world.