The method behind the unprecedented production of indicators of the presence of languages in the Internet

Abstract

Reliable and updated indicators of the presence of languages in the Internet are required to drive efficiently policies for languages, to forecast e-commerce market or to support further researches on the field of digital support of languages. This article presents a complete description of the methodological elements involved in the production of an unprecedented set of indicators of the presence in the Internet of the 329 languages with more than 1 million L1 speakers. A special emphasis is given to the treatment of the comprehensive set of biases involved in the process, either from the method or the various sources used in the modeling process. The biases related to other sources providing similar data are also discussed, and in particular, it is shown how the lack of consideration of the high level of multilingualism of the Web leads to a huge overestimation of the presence of English. The detailed list of sources is presented in the various annexes. For the first time in the history of the Internet, the production of indicators about virtual presence of a large set of languages could allow progress in the fields of economy of languages, cyber-geography of languages and language policies for multilingualism.