User talk:Bicolino34
Add topicСповіщення перекладу: Wikimedia Foundation Board of Trustees/Call for feedback: Board of Trustees elections
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Wikimedia Foundation Board of Trustees/Call for feedback: Board of Trustees elections доступна для перекладу, її можна перекласти тут:
Кінцевий термін для перекладу цієї сторінки – 2022-01-10 .
The Movement Strategy and Governance team is supporting this Call for Feedback. For widest outreach across the Wikiverse we kindly ask you to support this by helping us out with additional translations.
Postscriptum: During the last year many of you have helped us a lot to reach out to many different communities by translating in dozens of languages. We are utterly grateful for this - thanks to all of you for your ongoing support! You are the best!
Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 17:24, 7 January 2022 (UTC)
Сповіщення перекладу: Wikimedia Foundation Board of Trustees/Call for feedback: Board of Trustees elections/Call for Feedback about the Board of Trustees elections is now open/Short
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Wikimedia Foundation Board of Trustees/Call for feedback: Board of Trustees elections/Call for Feedback about the Board of Trustees elections is now open/Short доступна для перекладу, її можна перекласти тут:
Кінцевий термін для перекладу цієї сторінки – 2022-01-13 .
the Movement Strategy and Governance team is asking you for your help. We want to invite communities globally to a Call for Feedback about the Board of Trustees selection processes from 10 January to 7 February 2022.
We want to send out a message next Friday to all communities in as many languages as possible. It is only 130 words! Your help is very much appreciated!
Thank you for your time!Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 21:01, 11 January 2022 (UTC)
Сповіщення перекладу: Template:Education/News/Contents
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Template:Education/News/Contents доступна для перекладу, її можна перекласти тут:
Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 08:47, 22 January 2022 (UTC)
Сповіщення перекладу: Template:Education/News/Drafts
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Template:Education/News/Drafts доступна для перекладу, її можна перекласти тут:
Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 08:59, 22 January 2022 (UTC)
Сповіщення перекладу: Education/Newsletter/January 2022/Headlines
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Education/Newsletter/January 2022/Headlines доступна для перекладу, її можна перекласти тут:
Пріоритет цієї сторінки — високий. Кінцевий термін для перекладу цієї сторінки – 2022-01-24 .
Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 17:26, 22 January 2022 (UTC)
Сповіщення перекладу: Education/Newsletter/January 2022
[edit]Привіт, Bicolino34!
Ви отримали це повідомлення тому, що Ви зареєстровані на Meta як перекладач такими мовами: українська і російська. Сторінка Education/Newsletter/January 2022 доступна для перекладу, її можна перекласти тут:
Ваша допомога вітається. Перекладачі, як Ви, допомагають сайту Meta функціонувати як справді багатомовна спільнота.
Ви можете змінити свої налаштування сповіщень.
Дякуємо!
Координатори перекладів Meta, 17:29, 22 January 2022 (UTC)
Updates from Capacity Exchange (CapX)!
[edit]
Hello Bicolino34,
My name is Joris Darlington Quarshie, and I am an Outreach Facilitator for the Capacity Exchange (CapX) — a platform for finding and connecting with peers to exchange knowledge and skills on a global level.
CapX is a community-built tool, designed by and for the Wikimedia Movement, to make it easier for Wikimedians across the world to find and connect, collaborate, and exchange skills and knowledge with peers.
If you don't know it yet, I invite you to access capx.toolforge.org and create your user profile with your Wikimedia account. But, if you already have a profile, I invite you to check out the new features and possibilities we just launched. I am sure you will be delighted.
CapX's new version incorporates valuable community feedback gathered over the past months, including the Organization Profile - a feature for Wikimedia affiliates, initiatives, or community groups to create their own CapX presence (see here how to create yours).
Explore CapX's new features
[edit]→ Access: capx.toolforge.org
→ Watch the launch video:Meet the Capacity Exchange
→ Navigate CapX's User Guide and Frequently Asked Questions
→ Share your profile on Meta by adding the predefinition {{CapacityExchange}} to your userpage.
→ Visit our Meta Page for all documentation on the project
Please, let me know if you think your community would enjoy a presentation, workshop demonstration, or feedback session. I would be happy to support you on that. Just reach out, and we can schedule the best date and time together.
Best regards,
Joris Darlington Quarshie
Outreach Facilitator,
Capacity Exchange Project – Wikimedia Brasil
MediaWiki message delivery (talk) 14:39, 17 October 2025 (UTC)
Join us for “Many Tongues, One Movement: Voices Across Languages”!
[edit]
Hello Bicolino34,
We’re excited to invite you to an inspiring global virtual gathering: the first Capacity Exchange Translat-a-thon.
Together with Language Diversity Hub, the Capacity Exchange (CapX) team will host its first Translation Marathon dedicated to ensuring linguistic equity in access to this amazing tool aimed to connect Wikimedians.
If you enjoy contributing to Wikimedia projects through translating and adapting content into different languages, this event is for you! Join us in the celebration of the multilingual spirit of the Wikimedia Movement at an event where communities that contribute in diverse languages will be able to share local knowledge and collaborate across borders.
Many Tongues, One Movement: Voices Across Languages
- Date: December 6, 2025
- Time: 12 PM (UTC) - Check the event page for your local timezone
- Location: Online (Meta-Wiki + live session links)
If you can’t join the live event, you can still contribute to the translations! Edits will be counted for two weeks, until December 20th. And everyone who participates will receive a special badge to display on their CapX profiles.
Strengthen your collaboration through CapX
[edit]We invite you and your community to join the Capacity Exchange (CapX), a Wikimedia community-built platform for connecting, collaborating, and exchanging skills with peers across the movement.
CapX helps Wikimedians and organizations find each other, share expertise, and build stronger, more connected communities.
Whether you’re an individual contributor, a user group, a community initiative or an affiliate, CapX helps you grow through knowledge exchange.
More information
[edit]→ Explore the CapX platform: capx.toolforge.org
→ Read: User Guide & FAQ
→ Watch: Meet the Capacity Exchange video
→ Join our Telegram community chat: CapX Telegram Group
If your community, usergroup or affiliate would like to have a CapX organization profile, please reach out at capx@wmnobrasil.org, and we’d be delighted to support you.
With warm regards,
Joris Darlington Quarshie
Outreach Facilitator,
Capacity Exchange Project – Wikimedia Brasil
MediaWiki message delivery (talk) 14:55, 13 November 2025 (UTC)
Transkribus model
[edit]Hi,
I stumble upon https://app.transkribus.org/models/public/text/ukrainian-wikisource and if you'll allow me, I have so many questions ;) I'm wondering how you did it, if it was hard and if it was worth it (problems and benefits), etc. And mainly, should we try to replicate it for other Wikisources?
Cdlt, VIGNERON * discut. 20:57, 28 January 2026 (UTC)
- Hi,
- It was definitely worth it for Ukrainian community. Ukrainian language had a lot of different orthographies, some of which used a lot of symbols not present in modern day standard. Tesseract model was relatively okay for post 1920 texts of reasonable quality, but it failed miserably with older texts and lower quality (and even in best texts it tended to modernize orthography, for example replace соціяльний with соціальний). Google was about the same quality, but it also tended to mix Cyrillic and Latin symbols in the same word. There was also added complexity of fragments in other languages (mainly russian, Polish, German, Czech; it depends on type of work, where it was published and at what time).
- So when I stumbled upon news that Transkribus is being integrated into Wikisource I decided to try it out and make model for Ukrainian. Initial transcription for training was quite tedious since you need to mark up lines and add transcription for each line. You can do machine recognition of layout and text if there are already models available and you would only need to fix the mistakes. Unfortunately, there was no model for text recognition Printed Ukrainian at the time, so I had to manually type out transcription for each line (layout I recognized automatically). When I had about 20 pages (around 4-5 thousand words, I think), I trained basic text recognition model which sped up the process a little bit, but the quality was still awful. More or less reasonable quality was achieved around 20-30 thousand transcribed words and then process becomes much easier and faster. From that point you can just do automatic recognition and there will be only small errors when the type of pages is similar to your transcribed data. You can also quickly add more diversity to your dataset, so that it works better on wider variety. If your language already has some models then you can use them to streamline the process and it would be easier to start.
- All of the models I did were trained progressively, so I made some model, made it public for use in Wikisource and then after some usage in Wikisource noted scans where it performed badly, added those new scans for training and then made the next version of model. And so on.
- Timeline goes like this (names are from the Transkribus, they might be a little bit different at Wikisource for clarity reasons):
- Printed Ukrainian 20th century (September 2023)
- Ukrainian Wikisource (February 2024)
- Chuhaister (May 2025)
- Ruthenian Chuhaister (December 2025)
- The only problems I would note is that Transkribus has credit limits (I think, it was 50 or 100 each month for free, if more you need to pay; one credit equals 2 pages of printed text or 1 pages of handwritten) and some functionality is limited for non-paying users. If you have a subscription there are also models for region detection (if you want to train model for multi-column text) or table recognition. I worked with free version only and didn't need subscription since harder formatting wasn't my priority. So, my models don't quite work with even two-column text, but it can be solved for the time being by manually selecting column in Wikimedia OCR interface.
- There is an article on Wikimedia Ukraine site where you can learn more about our reasons. It's in Ukrainian, but I I think machine translation of LLMs would translate it well https://blog.wikimedia.org.ua/2025/11/06/from-tesseract-to-chuhaister/
- I thinks it is definitely worth trying for a lot of languages, especially if they are poorly supported by Tesseract and Google (since the models they provide are already falling behind). I think, you can train a reliable model for a language in a month or two if you spend more time than I did, but that's only if your language uses alphabet or syllabary. Language with hieroglyphics would be much harder (e.g. Chinese or Japanese). Bicolino34 (talk) 09:35, 29 January 2026 (UTC)
- Interresting, thank you (i come from french community ok wikisource). Sicarov (talk) 10:45, 30 January 2026 (UTC)