Translating telephone: problems and prospects

Summary form only given. Japan recently initiated a seven year $120 million project as the first phase toward developing a phone system in which a Japanese speaker can converse with say, an English speaker in real time. This requires solutions to a number of currently unsolved problems: (1) developing a speech recognition system capable of recognizing a large (possibly unlimited) vocabulary of spontaneous, unrehearsed, continuous speech, (2) developing a natural language translation system capable of dealing with ambiguity, nongrammaticality, and incomplete phrases, and (3) achieving natural sounding speech synthesis that preserves speaker characteristics