Good morning, Mr. Chairman, honourable members.
My name is Francine Kennedy, and I am the Chief Executive Officer of the Translation Bureau. With me this morning is Donald Barabé, Vice-President of Professional Services at the Translation Bureau.
Let me begin by thanking you for inviting me to come and talk to you about the usefulness and limitations of machine translation systems and the possibility of using them for the Job Bank.
Before I do so, with your permission I would like to take a few moments to outline the Translation Bureau's mission and give you a snapshot of our services and the various translation tools that we have. I will also give you a brief introduction to language technologies, their potential and their limitations. Lastly, I will speak to you about the Translation Bureau's contributions and recommendations with respect to the Job Bank.
The mission of the Translation Bureau is to support the Government of Canada in its efforts to provide services for and communicate to Canadians in the official language of their choice.
Since its creation in 1934, the Translation Bureau has been meeting the translation needs of federal government departments and agencies, the judiciary, both houses of Parliament, parliamentary committees, and the Prime Minister's Office. This mandate was broadened in 1987 to include vocal interpretation, sign language interpretation, and our terminology standardization program.
In 1995 the Translation Bureau became a special operating agency of Public Works and Government Services Canada. As a special operating agency, the Translation Bureau operates in a businesslike environment. Services are optional and on a cost-recovery basis, with the exception of services to Parliament and the terminology standardization program, which remain vote funded.
Federal institutions can therefore choose to do business with the bureau or to contract out their needs directly to the private sector in accordance with contracting rules. The decision is entirely up to the institution.
The Translation Bureau has 1,800 staff, including 1,200 translators, interpreters and terminologists. It has 60 points of service across the country from St. John's, Newfoundland to Vancouver, British Columbia. The Bureau provides services for over 150 departments, agencies and commissions. We translate in both official languages and in over 100 aboriginal and foreign languages. We also provide interpretation services in 40 languages.
We produce bilingual glossaries on a wide variety of subjects. For example, we recently produced one on diamond cutting, in cooperation with the Matane CEGEP, at the request of the Northwest Territories. We also have electronic terminology products, such as Termium, the Government of Canada's linguistic databank, which contains over 3.5 million terms. We also manage the Language Nook, a website that is available to all public servants to help them work in both official languages and provide them with a variety of electronic writing tools.
Last of all, the Translation Bureau supports the implementation of Canada's Official Languages Act and ensures that the federal government has the language services it needs, by supporting the development of the Canadian language industry, promoting the development and dissemination of standards in Canada and offering its specialists' support to other departments and agencies.
I will now address the topic of language technologies, their potential and their limitations. At the risk of disappointing you, I must start off by saying that, right at this moment, in spite of all the technological progress since the unveiling of the first automatic translation system in the United States in 1958, machines are not yet able to take the place of flesh-and-blood translators. Further progress would require artificial intelligence-based technology, the development of which remains a considerable way off. Translation is a complex mental process. Translators do not translate only words. They also translate ways of thinking, as well as cultural, social, intellectual, scientific, technical and other values — human values.
This is not to say that there is no place for machine translation or computer-assisted translation; on the contrary. In controlled contexts, machine translation can be an extremely useful tool that can increase the productivity of human translators. We speak from experience, as the Translation Bureau is in fact the oldest and one of the largest users of machine translation in the world.
We have been using a machine translation system for some 30 years to produce bilingual communications that are broadcast every day, come rain, come shine, going out as many as 250 times a day, 24 hours a day, 365 days a year, to the entire population of Canada from coast to coast to coast. As you might have guessed, I am referring to our weather reports, which we translate using a machine translation system. This entails the production of a substantial quantity of weather reports, that are translated into either French or English, depending on where the text comes from. A team of translators works on site to validate the translation, which involves reading and, where necessary, correcting the translation produced by the machine translation system.
Why, you may ask, does machine translation work well in this particular case? The key is in the writing. Weather reports are subject to controlled writing that uses a pre-determined format. The authors must, therefore, comply with a writing protocol. Controlled writing is the only possible way to use a machine translation system effectively.
Research continues and our technology watch team is monitoring technological developments in this area very closely. An interesting example is the recent development of systems based not on lexical data but on statistical sampling, which could improve machine translation significantly.
Organizers of the 2008 Olympic Games in Beijing gave an undertaking that a comprehensive multilingual information service would be available at the games at all times and in all locations. To achieve this, they will use state-of-the-art language technologies. Those in charge did, however, reach the conclusion that, while several technologies are mature enough to be integrated, others such as free text machine translation need to be combined with safer language transfer methods and must always be combined with the opportunity to fall back on human translation.
Other tools, such as translation memories, also provide effective solutions to the constantly increasing demand for translation, something which is not limited to the federal government or to Canada, but is a worldwide phenomenon. Translation memories make it possible to store sentences or parts of texts that have already been translated and to retrieve them for use in a new translation. Each source sentence and its translation has to be entered in the database. These systems are recommended in the case of texts that are very repetitive.
The Translation Bureau plays an active role in the area of technological tools. For example, we are present in all of the major international fora that deal with language technologies. We maintain close links and exchange opinions, advice and expertise with those who direct the largest translation services, including those of the European Community and United Nations. We are one of the founding members of the brand new Language Technologies Research Centre housed by the Université du Québec en Outaouais.
Let us go back to the Job Bank, however. Mr. Chair, while the Translation Bureau does not translate job offers, in recent years, it has worked with the Job Bank's team in a number of areas. During this time we have continually said that machine translation systems are still not sufficiently well developed to translate free text and that subsequent revision by a real person would be required.
We sympathize with our colleagues from the Job Bank, because the challenge they face is a substantial one. Of course the solution lies in technical improvements such as the ongoing enhancement of the dictionary, but of even greater importance is the use of menus and controlled writing, i.e., the elimination of free text.
While the solution may seem simple, its implementation may prove complicated. I can only imagine the numerous technical, budgetary and political constraints those in charge of the Job Bank have to contend with. This undoubtedly explains why the Job Bank's team followed up on a number of technical recommendations, but was reluctant to introduce a writing protocol and impose stringent limits on, or entirely eradicate, the use of free text in job announcements. We are continuing to hold discussions and are willing to offer our assistance.
Thank you, Mr. Chair. That ends my presentation.