Douglas Hospital in Verdun, Que., faces a typically Canadian problem: half of the staff and patients speak English and half speak French.
So when it came time to implement a new intranet strategy, the psychiatric facility decided to acquire technology that would automatically translate all internal documents between the two languages.
Daniel Sirois, service manager for management information systems at the hospital, conducted a study to determine whether a translation engine would be worthwhile before developing a corporate strategy for the hospital’s new intranet.
“We discovered that many users [translate documents] partially or with an external resource. It was done very individually.” This caused a delay in the availability of information as well as an administrative lag, he said.
The Domino-based intranet at Douglas, the first section of which is scheduled for use in mid-march, is being developed by Montreal-based Sonya Inc., which will also integrate the translation package.
Because the facility spans 37 different buildings, the biggest problem was keeping information current across all sites, said Charles-Jean Gollain, president of Sonya Inc. in Montreal.
“The first step is to implement the intranet so we can diffuse the information more quickly and making sure information is always available,” he said.
The idea behind the intranet is not to change the way people are currently working but to map existing processes to user needs, and then optimize those processes. This is one reason being able to read documents in two languages is so important, he said.
The translation gateway, Alis Translator for Lotus Domino (ATLD), developed by Montreal-based Alis Technologies Inc., embeds directly into the Domino intranet environment on the server end. It is a Unicode-based system so it will handle different character sets of various languages, said Steve Allan, senior director of Alis Technologies Inc. in Montreal. Developers creating notes applications use Lotus Script or Java and a set of function calls to translate content of their applications, he said.
“Within their application they can plug translation in anywhere that their business process requires.” It translates on demand or automatically, depending on how the user sets it up.
The program includes tools for language detection, text analysis for proper names, geographic locations and dates, then tags them so the translation engine can handle them properly.
According to Allen, the program uses structural analysis of the language, not word matches. “The resulting quality that you get out of the engines is obviously less than perfect – less than you would get from a human translator.”
Formal language usage gets the best results, he said. “If I’ve got something like a technical manual that tends to use a very formal sentence structure and usage, it will do a very good job on that. But if I’ve got something very casual, such as an e-mail where I’ve used all sorts of little ellipses and choppy messages, it has more trouble.”
Douglas Hospital is using the program primarily as a tool to help staff translate English documents to French and vice-versa in order to save time and administration costs. It will also help ensure quick memos and other information being pushed across the system will be understood by everyone.
“What happens is you get people who are not fully fluent, who are obliged by policy to do casual translations. So it’s a purely informational thing.”
Although the automatic translation saves time, human-based translation is still used for critical medical documents, Douglas’ Sirois said.
“If you know that the sender translated the message, you know it’s a machine-translation and that it’s not 100 per cent (accurate). Alis was very honest about that,” he said.
“But it’s a good solution…that’s very focused on the needs of the individual and respects the person’s choice of language.”