This paper deals with early Machine translation. It was mainly anchored in “the first mathematization of language” of the 1930s which was characterized by the rise of formalization as promoted by the School of Vienna as a common objective for every science. Thanks to computer programming, the interaction of syntax, logical mathematics, formal languages and algorithms became dynamic and formed the second mathematization of language, a common background of the two main personalities who promoted Machine Translation, Warren Weaver (1894-1978) and Yohoshua Bar-Hillel (1915-1975). I intend to address two European traditions comparatively, and examine how they dealt with the integration of machine translation into the language sciences. Early British machine translation experiments and early Russian experiments show similarities. While working completely separately, the British and the Russians developed machine translation methods based on semantic intermediary languages, anchored in their linguistic and philosophic specific traditions. I will examine which theoretical and methodological transfers had been necessary to carry out such types of integration.