This machine translation research is joint work of Ralph Grishman at New York University, Michiko Kosaka at Monmouth College and Virginia Teller at Hunter College of CUNY. Progress in machine translation has been hampered by the lack of clear development methodologies. The goal of this work is the creation of such a methodology. These investigators suggest an approach based on the cross linguistic sublanguage analysis: the systematic collection and correlation of the patterns of word usage in the source and target languages (in this case, Japanese and English) for a particular subject domain. These comparative linguistic studies will be based on an operator argument formalism, which in most cases reduces corresponding English and Japanese sentences to structurally identical representations. The results of these studies can then be used to create a translation system, consisting of an analysis phase (creating operator argument trees), a transfer phase (based on the sublanguage patterns), and a generation phase (to synthesize target language sentences). A prototype system will translate computer manuals from Japanese into English.
