The development and validation of language translators (e.g. port programs, language preprocessors, high-level software language compilers, etc.) are time-consuming and error-prone: language engineers need to master both the source and target languages' syntactic constructs; and most importantly their semantics. In this paper, we present an innovative approach for developing and validating such language translators based on two languages: With the first, we specify a language translation using a syntax-to-syntax mapping; and with the second, we define the semantics of both of the source and target languages. After showing how such specifications can be combined to validate and generate language translators automatically, we demonstrate the feasibility of the approach on a particular modelling language translation.