User:CFCF/botspecs

Two example articles we've translated into Khmer and Yoruba

The WikiProject Medicine Translation Taskforce is a major project which aims to translate high quality articles on health topics into as many languages as possible. We've so far managed to translate over 1300 articles into 116 different languages! We're especially proud of our work translating information into smaller African, Asian and South American languages which have very limited content even outside of Wikipedia. Part of the project is a collaboration between Translators without Borders (TwB) and Wikiproject medicine. We use their translation platform extensively, and to keep everything running smoothly we've handled all tracking and translation requests manually. With the amount of participants we have today this no longer works, and has become a major stumbling block for our future expansion.

Our translations start off as articles on the English Wikipedia and are passed through a number of web-interfaces in order to process our translations.

1. Google-docs: [1]

This is where we keep track of everything: articles that are to be translated, are translated, and live. It contains the links to our translations as well as the links to the articles on Wikipedia.

  2. Translators without Borders Translation Workspace: [2]

This is where We use this ProZ-based platform to send out translation orders to the Translators without Borders team of volunteers. It comes with an API detailed at [3].

  3. Wikipedia

This is where we put our translations, and where all our hard work ends up. Getting content on to smaller language Wikipedias is our main goal.

What is the problem and our solution

We want to automate the notifications process so that volunteers can focus on the final goal of integration without having to continually update and notify manually at every step.

Our solution involves four components: 1) a translation request system where an editor can mark an article that is ready to be translated in the google spreadsheet, and then request automatically posts as TwB work-order using their API; 2) a system where the editor who ordered the translation gets a message on their Wikipedia talk page that it was translated and is ready to be integrated into Wikipedia; 3) a system where the Google tracking pages are auto-updated once a translation is first requested and then again it is marked as translated.

Bot documentation

What should the bot do?

The bot is meant to send out translation work-orders to TwBs upon request by a Wikipedia user. Once the translation is finished the users is informed of this so that they can go ahead and integrate the article. Preferably hosted at https://tools.wmflabs.org/
To post we want users to register: username & mainwiki (mainwiki can be a drop-down menu)

Creating orders

1. Take the article and post it through the TwB API
   Frontend:
   1. Drop down menu of preparedarticles from Simplified list (available in at google drive: https://drive.google.com/folderview?id=0ByM3vikRdwyzYTlYaG92Q2w3VWc&usp=sharing under folder 1-50, 51-100 – it should be easy to add more such as 101-150 etc.. format should be documented)
   2. Checklist including all languages available on TwB platform

Backend:

Take prepared file from google docs & post to TwB (needs login & API access)
      User': Jmh649
     Password: wikipedia

Traditional create page[4]
API information: [5]

Fill for each article:

Title:Article title
Specialization:General
Source language:English
Deadline: 1 month
Client order number:parameter currently unused
Source file:taken from google docs in format "##,Articletitle MMMYYYY En.docx (or .doc)"
Supporting materials: [6] [7]
Target language:as chosen by user
Deadline flexibility: Flexible
Job invitations: Automatically

Create order

Each created order (1 per/language) gives rise to a Job detail page such as this one for Leishmaniasis into Dutch

This page is linked at the google spreadsheet tracking page under the correct language, article and TWB and under correct status. As soon as the order has been sent it should be linked int the light blue format detailed here [8] with the caption "Sourc"

Scanning created orders

Once a day the bot should scan the orders which are marked as "Sourcing" (Sourc) or "Translating" (Transl) on the tracking pages [9] [10] [11] [12] [13] [14]. (It should be easy to add more tracking pages here – preferably by just adding a link to a config file).

The bot double checks these against the TwB work order page and updates colors according to [15] in the spreadsheets.

Once a work order reaches the Complete (compl) stage or Failed (Failed) stage the user who posted the work order should be contacted through a mass-message at their home-wiki. This works through the meta MassMessage function detailed there. Message should read

The article on "ARTICLENAME" you requested for translation has now been translated into "TARGETLANGUAGE" and is available at [TwB link (http)]. Once you have integrated the article please update the [https://docs.google.com/spreadsheets/d/1cb80jUe-tObwbTo-o4hh2IpcQHSv1TAJh-8vuniNsCs/edit#gid=1780120699 tracking page]<br>
~~~~