Abstract
Working from a dataset of 118 billion messages running from the start of 2009to the end of 2019, we identify and explore the relative daily use of over 150languages on Twitter. We find that eight languages comprise 80% of all tweets,with English, Japanese, Spanish, and Portuguese being the most dominant. Toquantify social spreading in each language over time, we compute the 'contagionratio': The balance of retweets to organic messages. We find that for the mostcommon languages on Twitter there is a growing tendency, though not universal,to retweet rather than share new content. By the end of 2019, the contagionratios for half of the top 30 languages, including English and Spanish, hadreached above 1---the naive contagion threshold. In 2019, the top 5 languageswith the highest average daily ratios were, in order, Thai (7.3), Hindi, Tamil,Urdu, and Catalan, while the bottom 5 were Russian, Swedish, Esperanto,Cebuano, and Finnish (0.26). Further, we show that over time, the contagionratios for most common languages are growing more strongly than those of rarelanguages.