From user feedback: "In English for representing one thousand we write like this 1,000 (with a coma) but in Catalan we use a dot, 1.000 . For decimals we would write 0.54 in English but we use 0,54 in Catalan."

Ideally this should be handled by the MT engine, but if it's not done there, maybe we can do something ourselves by identifying the numbers in the text and using the existing CLDR-based functions.