ENGLISH-FRENCH, FRENCH-ENGLISH MACHINE TRANSLATION I received some material from Dave Sutherland at the National Capital Freenet. In it was an article by Dorothy Phillips, Thomas Whalen, et al on "The Changing Face of Broadcasting..." I fully concur with the sentiments expressed in that article. I have been pursuing a similar objective, but in the area of inter-lingual communications. I strongly feel, that with our current constitutional problems we need communication tools to allow citizens of this country, from different linguistic backgrounds, to speak to each other in their own native tongues. Only through communications and understanding of each other's view points can we ever hope to resolve the linguistic problems that plague this great country of ours. To that end, I have been doing considerable research into machine translation systems for mono-lingual users on electronic networks like FreeNet. The concept I have in mind, which I hope to turn into a VISION 2000 project, is the development of a machine translation system for mono-lingual users that will allow members of the public, of one language group, to access networks like FreeNet and converse with other individuals who speak a different language. Bulletin boards or conferences could be set up where members of the community could chat about their different linguistic viewpoints on a range of topics, including the constitution, sports (the Leafs vs Les Habs), culture, etc. Electronic pen-pals could be set up for school kids so that they can communicate to other kids in a different language. The possibilities are endless. I have recently attended the TMI-92 conference in Montreal and I have been in correspondence with the MT people at CWARC regarding this subject. Related work is already being done in this area at CompuServe and Carnegie- Mellon. There are significant restrictions to setting up such a service, due to the limitations of the technology, but I still think there is a real potential here to expand the communications horizon of all Canadians and develop new machine translation tools for the Canadian communications industry. To my mind the NC FreeNet would be an ideal test bed for the necessary research and development in this area. There could be many technological spin-offs for Canadian industry in MT (Machine Translation) products and services as well as providing a valuable service to the community at large. The major problem with MT today is its very limited use in general subject areas. However, moderate success is being achieved in what is called restricted domain authoring systems. Machine translation on these systems works by restricting the subject area, topic, vocabulary and semantic structure of the author's document. They are ideally suited for electronic networks and computers and not pre-written documents. Carnegie-Mellon is doing pioneering work in this area with their recent $5,000,000 contract to translate all of Caterpillar Tractor's service manuals. The important thing to remember here, is that they are not translating existing manuals, but they are setting up authoring systems for writers of new manuals so that they will be written in a style that will make it easy for a computer to translate. Simply stated, the computer system prompts the author of a document whenever they are writing in style that is too convoluted for the computer to understand. Pop down menus, like those in a spelling checker appear whenever the computer cannot resolve ambiguities in meaning or vocabulary. Similar application work has been done at CWARC with their CRITTER translation system. My personal objective is to see if we can extend this development work in restricted domain authoring systems to a larger audience. E-mail users are generally much more tolerant of grammatical errors and so 100% accuracy is not a requirement. Those familiar with USENET know that there are hundreds of very technical and restricted domain news groups. Many would be ideal candidates for MT usage. Other areas for this type of MT is EDI (Electronic Data Interchange) which is a very restricted domain used for commercial electronic invoicing, ordering, etc. To my mind, electronic pen paling would be the first ideal application. Children's semantic structure is very simple and their vocabularies are limited compared to those of most adults. I also think this would be fertile ground for starting the dialogue between the linguistic dualities in Canada. If you think there is merit in this proposal, I would love to hear from you. As I mentioned above, I have a personal interest in a project of this type. Despite my surname, I am essentially unilingual English. I have two brothers, one who is unilingual French (and a avowed separatist) and another who is fluently bilingual. Our family is sort of a microcosm of Canada. But unlike most Canadian families, all three of us have computers and e-mail. Need I say more. Regards Bill St. Arnaud, P.Eng. Project Officer Vision 2000 567-2000 email: wcsv2k@carleton.ca VISION 2000 is a private and public partnership formed in 1989. Made up of 38 key Canadian Communications and Information Technology companies, research organizations, academia and governments. Vision 2000 aims are to accelerate innovation and foster collaboration in "personal" communications and informaioon technology R&D, as well as to introduce new Canadian products and services to world markets.