Canada, Cats & Dogs, and Machine Translation
Bilingualism in Canada is a big deal. Back in 1961, before the modern Canadian flag was born, and back when it was still called the Dominion of Canada, the percentage of bilingual Canadians, those who were fluent in both English and French, was only 12.2%. Less than one person in eight.
By 2011, this number increased to 17.5%, or more than one in six. In Quebec, with its concentration of French speakers, the bilingual percentage rose from 25% in 1961 to 42.6% in 2011.
Back in 1961, the population of Canada was only 18 million. The population of Canada has nearly doubled since then to over 35 million, of whom nearly 6 million Canadian citizens are bilingual, and more than a million others speak French natively. With English and French both official languages of equal status, this puts a significant requirement on government officials to understand and communicate effectively with their constituents.
Within the Canadian government, the Translation Bureau (TB) is chartered to bridge such linguistic gaps, providing both human and machine translation (MT) services. Founded in 1934, many question whether it is keeping pace with the rate of technological change and the amount of work that’s required. It recently had its staff cut from 2,000 to 1,300 over five years, with much of its work now outsourced to the private sector.
What Canadian civil servants are doing in response should not be surprising. Like most people, they are using Google Translate en masse, at a rate of over a million times each week, according to the TB’s CEO, Donna Achimov.
To help bolster its MT offerings, in April 2016 the Translation Bureau, working in conjunction with the National Research Council, plans to roll out a new system to 350,000 Canadian government workers. The new system, called Portage Statistical Machine Translation, has been under pilot with multiple governmental ministries since last summer.
This is where the controversy lies. Critics question the quality. The results have been described as “quirky” and “clumsy.” For instance, the French version of the system’s description, «nouvel outil de raduction automatique», was rendered “new tool machine translation,” rather than the more natural English construction “new machine translation tool.” It also translated the idiomatic English expression “It’s raining cats and dogs,” to «La pluie, les chats et les chiens.» (“The rain, the cats and the dogs.”) For comparison, Google Translate renders the same phrase as «Il pleut des seaux d’eau» (“It is raining buckets of water.”)
The Canadian Association of Professional Employees (CAPE), who represents the Translation Bureau’s workers, told the CBC that Portage system might be less capable than Google Translate because it will not have the same broad population of people suggesting improvements to the translation. The union’s many professional translators are leery of seeing their department’s reputation tarnished with results portrayed as amateurish or awkward.
The tool is part of a hybrid plan for human and computer-aided translation. Portage results are never meant to go directly to Canadian public. Information that needs to be sent out to the public should be linked directly over to a human translator.
The main advantages cited for the Translation Bureau hosting its own internal translation system is for building up its own government-specific translation memory, and for security and confidentiality, since the results would stay internal to the Canadian government, and not processed on public Google servers in the United States.
One thing is for sure, we’ll be watching the rollout in April with great interest! Others will be as well, since the NRC is looking to license the Portage system commercially through Terminotix.
How about your own organization? How are you balancing your work between machine and human translators? Send us an email at projects@e2f.com. We’d love to hear your stories!