SYSTRAN, whose name stands for “SYStem and TRANslation,” was founded in 1968 by Dr. Peter Toma, a Hungarian-born computer scientist who believed that better communication between nations could help secure world peace. Working in California during the Cold War, Toma built one of the first practical machine translation systems by cooperating with the US Air Force to translate Russian scientific and technical documents into English. SYSTRAN appeared just two years after the 1966 ALPAC report had concluded there was no near-term prospect of useful machine translation and effectively dried up US government funding for the field.
SYSTRAN was a rule-based system. Rather than learning from data, it relied on hand-built bilingual dictionaries and large sets of linguistic rules to analyze a source sentence and generate a target sentence, the dominant design approach to machine translation for decades. Toma’s Russian-to-English software ran for the Air Force’s Foreign Technology Division throughout this period, processing large volumes of Soviet documents that human translators could not keep up with.
The system reached a public milestone in 1975, when, by SYSTRAN’s own account, it “translated instructions from Russian into English for the APOLLO-SOYOUZ mission,” the joint US-Soviet spaceflight that required American and Soviet crews to work from shared procedures. To support the project Toma added an English-to-Russian capability, creating an early two-way translation package at a time when many linguists doubted that translating out of English could be automated at all.
SYSTRAN went on to power the European Commission’s translation work and, decades later, the first version of Babel Fish and early Google language tools, making it one of the longest-running brands in the history of language technology. Its survival across the rule-based, statistical, and neural eras of machine translation makes it a useful thread connecting the field’s earliest Cold War experiments to today’s neural systems.