Wikipedia as a translation machine? Sounds strange at first. That’s what Microsoft Bing Translator, Google Translator, DeepL or the many professional translators and agencies like us are for.

The idea is quite different. Denny Vrandečić is working on a universal language of facts.

The “Wikipedia” project now includes more than 50 million articles in a wide variety of languages. Many of the articles are not simple translations from other languages, but articles created in the respective language. Thus, they are not “globally” available or readable, but only in the source language. A machine translation is usually difficult, because it concerns languages, for which too few language data are available in order to generate a qualitatively sufficient machine translation. Or they are dealing with very specialized subject matter.

The principle behind “Abstract Wikipedia” can be illustrated with a simple example. Instead of talking about “half” in a standard text, one uses, for example, the mathematical markup “50%”. This can then be replaced by terms like “half” or “la moitié” when rendering the respective output language.

This idea is also applicable to the documentation of, for example, machines or software products and makes translation easier – whether by professional translators or just machine translation.

And finally, Vrandečić also makes an interesting statement about his native language, Croatian: “Besides, the language has changed a lot in the last 20 years, this has political motivations, many terms are redefined today to distinguish Croatian more from Serbian.”


Wikipedia wird 20: »Wenn Google das Projekt angehen würde, das würde glorios scheitern« – DER SPIEGEL