Last week you may have heard, or even seen, that Twitter has been trialling automated translation of tweets from certain individuals in Egypt. Following the recent political upheaval, the former president Mohammed Morsi, opposition leader Mohammed ElBaradei and Arab Spring activist Wael Ghonim have all had tweets made available to non-Arabic speaking readers.
It is very much at the experimental stage, with Twitter not officially launching the service. The tool utilises Microsoft’s Bing translator to automatically generate tweets in other languages.
Given the inherent unreliability of these web based translation services, it does raise the question of whether machine translation should be used for applications like this. On the one hand it is incredibly efficient in making your message accessible to as many people as possible with instantaneous results. However, given the recent increase in criminal prosecutions targeted at individuals who have had their tweets misunderstood or just plain misuse the service, is it sensible to introduce another point of potential failure? Intentional misuse aside, someone using the system, is in effect relying on a machine to understand the context of the original message and correctly convey the language used so that it will not be misinterpreted. As advanced as translation systems are becoming, it is asking a huge amount to expect things not to go wrong, especially in a sphere that has such potential for volatile reaction.
If you think this is overblown, just take a look at this story from last year in the Guardian, focused on a tweet by Tory MP Adrian Burley. Regardless of the content or intention – it is proof that an ill-considered tweet can have serious ramifications even before introducing the murky reliability of machine translation. While most of us accept that the results from these systems are to be taken with a pinch of salt (or even a few gallons), as they become more prolific and enter more and more systems they can often assume a false notion of reliability in the public psyche.
This level of consumer reliance is a long way off for machine translation, however when/if it does begin to become second nature to trust the results, this is when the problems will arise. Sat Navs are a prime example – people start to ignore what their surroundings indicate and follow the computer generated instructions without question, most of the time it works fine – but there’s always a few cases a year where someone drives into a lake –in the end you can’t blame the box, you just have to keep an eye on what you’re doing.
What do you think?