Is machine translation still a dirty word?
There are few things that bring professional translators more joy than seeing examples of translations that have completely missed the mark and ended up as gibberish. However, is the tide turning in the relationship between human and machine translators, and could the two work together in harmony one day?
A thorny issue
This was the topic discussed by Jost Zetzsche and Charlotte Brasler for an article in last month’s ATA Chronicle, published by the American Translators Association. The pair acknowledged that while they are not averse to using technology like translation environment tools to make translating documents more efficient, machine translation is something that is often scorned upon.
However, in order to form an accurate opinion on this issue, translators need to familiarise themselves with what exactly machine translation (MT) is, how it works, what versions are available and how the technology is developing, rather than basing all their opinions on the free, online versions. So, Brasler and Zetzsche have admitted they were glad the 2012 ATA Annual Conference took place in San Diego – the same place as the Association for Machine Translation in the Americas conference. The close proximity gave professionals working in each sector the chance to learn about the other’s speciality.
“Most translators do not know much about MT – aside from the same-old-same-old jokes about the silly mistakes,” Zetzsche said. Of course, it’s little wonder when machine translations prove so hilarious – often causing blushes for people and business that rely on them.
Bad Translator is an Ackuna tool developed by Translation Cloud. Input up to 250 characters of text and the tool will run the content through anywhere between eight and 35 translations in and out of different languages using free machine translators like Bing, before giving you the end result in English. The finished translation is often about as far from the original text as you could imagine.
For instance, we took an extract from the Queen’s 2012 Christmas speech and ran it through eight translations. This was the result:
Original: “On the barges and the bridges and the banks of the river there were people who had taken their places to cheer through the mist, undaunted by the rain.”
Translation: “The comfortable boat and bridge and the River namestili to support him in the fog, the rain.”
Even running it through Google Translate once (translating it to simplified Chinese and back to English) the result is far from accurate: “Barges and bridges, river banks, where people take their cheering through the mist, without fear of rain.”
If you were in charge of translating your business’ website or promotional materials such as brochures, imagine what your customers would make of it if faced with this gobbledygook. They might find it funny but, equally, they could find it offensive as it suggests you have no real interest in communicating with them. Either way, they won’t understand what it is you’re trying to tell them and as a result your message will be lost.
This is why you should never rely on free software or internet apps alone to translate any document you plan for someone else to read. However, based on this, is it right that professional linguists continue to dismiss MT out of hand?
Not according to Brasler, who said that translators need to understand the ins and outs of machine translators, such as what languages they work best on, what the Bilingual Evaluation Understudy score is, and what the differences are between the rule-based, statistical and hybrid tools available. “When translators can speak intelligently on these subjects, we can engage in a much more professional dialogue with ‘the other side’ – the machine translation community,” she explained.
Rise of the machine translators
Brasler warned that the biggest mistake human translators can make is either to dismiss the machines by saying “machine translation is just Google Translate”, or to ignore it for fear they will be replaced. Instead, linguists should embrace the change in order to make the most of it. Yet there is just as much diversity within the human translator industry as the MTs, and the work a translator does can affect their opinion. Linguists range from established professionals to the recent graduates who feel they will be in direct competition with the machines.
Collaboration between humans and the machines could be the answer. Valarie Badame, marketing manager at Milengo, wrote in the company’s blog last year that just a decade ago translators viewed translation memory as the “death knell” for their profession. However, today the technology is widely regarded as a useful, and in some cases “essential”, tool that allows individuals and small groups to take on huge projects.
Language Insight has written before about how businesses are increasingly realising the value of translating and localising their content in order to break into new foreign markets. Indeed, it has never been easier to secure business overseas, thanks to the internet and the growing variety of digital communication. To take advantage of this, enterprises all over the world are seeking to get everything from their websites to their slogans translated. With this comes a huge volume of work for professional linguists to get through.
Badame noted that a few years ago numerous teams of linguists would have to be assembled to tackle projects involving a high volume of text and a large number of languages. However, thanks to machine translation and memories, these tasks can now be taken on efficiently by smaller teams. Yet there are many who would argue that this is a sacrifice of quality in favour of quantity.
A collaborative future?
“In order for this to be truly successful we need to be realistic and accept that in its current state, MT is not, and will not in the near future, be able to reproduce the linguistic nuances of a human being,” Badame explained. She added that even when the technology improves, rather than put human translators’ jobs at risk, the use of MT will mean there is a greater need for human specialists, as well as experts in the specific subject matter, to correct and polish the translation produced by the machine. It will also take a qualified translator to review the finished product and ensure it fits the brief.
It is also important to remember that there will always be jobs MT simply is not suitable for. While large-scale projects that involve a lot of repetition such as market research questionnaires could be covered by MT, human translators will always be chosen for creative, promotional and editorial work, Badame pointed out. She concluded that one day soon MT will be seen “as another tool in the arsenal of an advanced translation strategy”, rather than a threat.
It is a point Zetzsche appeared to agree with, as he claimed the stigma attached to using MT is “gradually going away”. Meanwhile, Brasler predicted: “We can expect to see exciting developments in the coming years as attitudes toward MT and confidentiality change.”
That’s what the experts think, but what’s your opinion on the role machine translation plays in the translation process as a whole? Share your thoughts below.