We believe in radical transparency and measurability, so we publically disclose the details of our internal processes and data. If we ever feel uncomfortable about publishing something, we see it as an opportunity for us to question whether our processes are the right ones.
Our vision and strategy are described on the homepage. Here, we go a bit further to outline the key principles of our ways of working. The more detailed internal process that our translators follow can be found here. The articles that we published in English are listed here, and our work-in-progress tracker (including translations published into other languages) is accessible here.
Prioritization of articles to translate
Our translators receive a short list of articles to translate, which they can pick from freely. When deciding which articles should be on the list, we follow a value maximization principle based on 4 primary criteria:
- Quality. We prioritize articles that are already well-written and comprehensive in the source language.
- Gap in the English Wikipedia. We prioritize articles that do not yet exist in English, or have barely been started.
- Value for society. Articles on important topics and/or likely to generate the most interest by readers are prioritized.
- Personal interest. To the extent that it is compatible with our other principles, our translators have full freedom to pick the topics that interest them the most. If articles that interest them are missing, they are encouraged to suggest new ones.
Before being translated, articles are manually assessed by our team of volunteers. To help them with the task, tools were developed to automatically identify articles that require translation.
Content diversity
An additional soft criterion that we consider is content diversity. It is not used for the prioritization of individual articles but is a dimension we look at when evaluating our overall impact.
While Wikipedia strives for a neutral point of view, most of its editors share social and cultural characteristics, causing systematic bias and imbalanced coverage of topics. Common characteristics of the “average Wikipedian” include “white”, “male”, “native English speaker”, and “from the Northern Hemisphere”.
Translating articles originally written in languages other than English helps reduce this bias. They often cover topics neglected by anglophones (e.g., Asian poets, South American mountains) or by male contributors (e.g., Female scientists), and introduce new points of view (e.g., they are more likely to quote research published by researchers from the Southern hemisphere).
Translation work
Once a translator has picked an article, (s)he first performs a few checks to ensure that very similar content doesn’t already exist in English under a different name, and identify related articles that will need to link to it. The real work can then start.
We primarily leverage Deepl, a machine learning-based translation tool, for the initial draft. The translator then reviews each sentence in detail to correct mistakes, improve the grammar, and adjust word choices based on the context of the article.
The next step is formatting. This includes adding links, images, tables, sources, etc. to make the article fit the Wikipedia style. This step is time-consuming and cannot yet be automated because each Wikipedia language uses different templates and tag structures (we’re closely monitoring the evolution of Large Language Models, and may increase automation levels once these are mature enough).
Once the article is ready, the translator publishes it under his own Wikipedia account. It is then public, and the Wikipedia community can freely make edits and corrections. But our work doesn’t stop there, as translators are also responsible for maintaining the articles they create (e.g., in case they receive feedback).
Translators selection and compensation
To maximize the impact of our limited funds, we primarily award grants to translators based in countries with low costs of living. These countries also tend to be underrepresented among Wikipedia editors, which also helps us decrease Wikipedia’s systematic bias.
Open positions are posted on LinkedIn and usually filled within less than 24h due to the high level of interest. Our translators work as freelancers and send us monthly “bills” for the hours they have performed. They are required to work an average of 20-40h per week to continue qualifying for OKA grants but are allowed to perform any other activity they want outside of the work they do for OKA, and have complete flexibility in their working hours. Most OKA grant recipients work part-time.
Administrative tasks for the management of the association are exclusively performed by volunteers; OKA doesn’t have any employees. Our statuses explicitly forbids members of the association to be remunerated by OKA.
Funding sources
Currently, the majority of our funding comes from our founders and from individuals, but we also accept donations from other charities, government entities, and companies.
If they wish to, our donors can broadly specify the area in which they want their money to be spent. For example, a local government may want its donation dedicated to the translation of articles related to their region. A company may wish us to focus on articles related to their industry.
But they are not allowed to influence the work of our editors under any circumstances. Our editors retain complete editorial freedom and are asked to ensure objectivity.
Legal structure
We are an association based in St-Légier, Switzerland (UID CHE-461.288.884). Our statuses are available here (in French). We are recognized as a tax-exempt organization with a public utility purpose.
Individuals donating at least 5000 CHF per year are eligible to become members of the association and to participate in the annual assembly.