The first steps in developing machine translation of patents

The TSNIIPI MT experimental system, developed in 1963–1966, was focused on the translation of publications of the US weekly “Official Gazette, specifically the first paragraphs of patent claims. These claims are characterized by an abundance of difficult to grasp multicomponent terminological combinations and by a specific syntactic structure of unusually long sentences containing up to several hundred words. The system's algorithm performed the segmentation of the English text, and the identification and structural analysis of multicomponent word groups necessary to synthesize the corresponding Russian equivalents.

► Context of present day patent machine translation systems.
► Historical perspective of the experimental MT system (TSNIIPI) in Russia in the 1960s.
► Segmentation and analysis of patent claim structures in TSNIIPI explained.
► Conclusion: current MT systems and TSNIIPI principles could usefully be combined.