A new approach to the translating telephone

MT Summit VII Sept. 1999 A New Approach to the Translating Telephone Robert Frederking Christopher Hogan Alexander Rudnicky Language Technologies Institute Carnegie Mellon University 5000 Forbes Avenue Pittsburgh, PA 15217 USA Abstract The Translating Telephone has been a major goal of speech translation for many years. Previous approaches have attempted to work from limited- domain, fully-automatic translation towards broad- coverage, fully-automatic translation. We are approaching the problem from a different direction: starting with a broad-coverage but not fully- automatic system, and working towards full automation. We believe that working in this direction will provide us with better feedback, by observing users and collecting language data under realistic conditions, and thus may allow more rapid progress towards the same ultimate goal. Our initial approach relies on the wide-spread availability of Internet connections and web browsers to provide a user interface. We describe our initial work, which is an extension of the Diplomat wearable speech translator. 1 Introduction The Translating Telephone has been the goal of several major speech-to-speech translation projects, such as those at ATR (Moritomi et al. 1993) in Japan, the JANUS group at Carnegie Mellon University (Waibel et al. 1991, Woszczyna et al. 1994), and others in the C-STAR con- sortium. While there have been numerous significant accomplishments in these projects over a number of years, the ultimate goal of a useful Translating Telephone still seems remote. We in the Diplomat project (Frederking et al. 1997, Frederking et al. 1999) have therefore decided to attack the problem from a new direc- tion, to determine whether more rapid progress can be made. Instead of gradually extending a limited-domain, fully-automatic system, we will begin with a broad- coverage, human-aided speech-to-speech translation sys- tem, and attempt to move towards full automation. The rest of Section 1 further describes our strategic view of machine translation research, and our previous work in this particular strategic direction. Section 2 pres- ents the current state of the Diplomat system, which pro- vides the foundation for this new thrust. Section 3 then describes the surprisingly minor extensions required to transform Diplomat into a Translating Telephone. We conclude in Section 4. 1.1 Trade-offs in current MT In order to understand the choice between these two strategic approaches, one must understand the trade-offs involved in the current state of the art in machine transla- tion (MT). As we see it, the ultimate goal of MT is a system that is: 1. Fully-automatic (no human intervention required) 2. General purpose (not limited-domain) 3. High-quality (the output is essentially as good as hu- man-produced) It is currently possible to achieve any two of these three objectives in a single system, using different state-of-the- art MT techniques; but achieving all three at once is still well beyond current capabilities. Seen from the most strategic level, one can therefore envision working from an initial system that satisfies any two of these goals, and striving to meet the third. The question that needs to be answered is which characteris- tics to begin with, and …

Content truncated to 500 words