Localization and internationalization facilities exist for applications, often as library functions (e.g. Posix gettext).
However, it makes much less sense to localize the source code of programs and scripts (e.g. by changing keywords of programming languages like
si in French....) because the meaning of a program is also conveyed by the identifier names and the comments.
Automatic and reliable (and faithful) translations of such names and comments is IMHO beyond the state of the art.
And I believe it would be simpler to have the machine program itself, i.e. synthetize its own code, instead of translating programs to be humanly understandable by other cultures. Look at Artificial General Intelligence and e.g. J.Pitrat's blog.
In practice, developers of software to be worked on by some international team (e.g. free software projects) should agree on some human language (often some form of English) and on some coding conventions or coding styles.
Some language don't have keywords (e.g. APL or even PL/1 where the same name
IF can have both a role of keyword and a role of identifier, so that
IF IF=THEN THEN; is a valid but cryptic PL/1 statement), but they do have identifiers and developers do give meaning names in identifiers for their own culture. Translating these automatically is not realistic.
Some very few publications mention using natural language processing techniques on comments and identifiers in static analysis of source code.