Tuesday, 22 November 2011

Google Translate: friend or foe?

Are you a language teacher? If so, what are your views on online translation tools?

Google Translate  is probably the most widely used online translation tool, but there are others that will also do the job. Several are listed in ICT4LT Module 3.5, Section 3 (Machine Translation). Such tools have been the bane of language teachers’ lives ever since they became widely available on the Web. The teacher sets a text to be translated for homework and the students use Google Translate to do the job, thus saving themselves work and driving their teacher mad when they turn in a piece of work that is full of mistakes that reveal clearly that an automatic translation tool has been used. Or the teacher may ask the students to produce an original composition in a foreign language - so they type it out in English and paste it into Google Translate. Again, the output is full of mistakes but often of a different kind, for example the students may be using constructions in English that are way beyond what they would be capable of using in the foreign language. And many mistakes made by Google Translate are made solely because the source text is incorrect. If you write "I should of thought" (yes, it's a common mistake!) instead of  "I should have thought" then Google Translate's output is wrong. But it translates "I should have thought" correctly into German as "ich h├Ątte gedacht". Thinking back to my early experiences with Machine Translation (MT) in the 1980s, I remember a company (Perkins Engines) that used the Weidner MT system first training its employees to write correct, unambiguous English so that the system could handle the texts more easily – in other words, anticipating potential errors that could be made.

Right now it's not too difficult to spot that Google Translate has been used to produce a text in a foreign language, but a few years ago Google began using a different translation engine that uses a so-called Statistical Machine Translation (SMT) approach. Now Google Translate begins by examining and comparing massive corpora of texts on the Web that have already been translated by human beings. It looks for matches between source and target texts and works out which translations are likely to be the most accurate. This YouTube video, Inside Google Translate, explains how it works. As more and more corpora are added to the Web this means that Google Translate will keep improving until it reaches a point where it will be very difficult to tell that a machine has done the translation. I remember early MT tools translating "How are you?" into German as "Wie sind Sie?" Now Google Translate gets it right: "Wie geht es Ihnen?" You can also click on the words in the translated text to hear how they are pronounced.
So Google Translate is no longer the crude tool that it used to be. Besides using a much more sophisticated and accurate translation engine, it also offers the possibility of interaction. When the translated text appears you can hover your mouse over the text and ask Google Translate to suggest alternative renderings if you don't accept what it offers as the first choice. These may be different vocabulary items, different tenses, different case endings in German, etc. You can also rearrange the word order. Thus you can edit the text until you are satisfied with it – and then you can copy and paste the text into Microsoft Word and edit it further using the inbuilt foreign-language spell checkers, grammar checkers and thesauruses. Having said that, I am in no doubt that most students would just accept what Google Translate offers as the first choice and hope for the best. But a clever student would investigate Google Translate's new features and produce quite an acceptable translation that does not have the obvious hallmarks of being translated by machine. So what is the solution if students cannot be persuaded not to use Google Translate?
  • Do you punish your students for cheating?
  • Do you hand back their work and tell them to do it again without using Google Translate?
  • Or maybe you warn your students that you have already run the text through Google Translate and that if you find any examples  of the same incorrect phrases being used in their work then they will score zero.
  • You could also exploit the mistakes that Google Translate makes by displaying them on a big screen to the whole class and showing your students how ridiculous they are. At the same time you could use the output of Google Translate to raise your students’ linguistic awareness. Ask your students to spot the mistakes and explain why they have been made – e.g. parsing like as a verb rather than a preposition.
But perhaps the time has come to admit defeat and to set different types of tasks for homework. A blog posting by Naomi Ganin Epstein, headed If Google is translating then I’ll start revamping, is worth looking at. She suggests setting a number of different types of assignments for homework that get round the problem of students using Google Translate.

Let’s face it, automatic translation tools have been around for a long time and they are here to stay. The European Commission makes extensive use of so-called Translation Memory (TM) systems. These produce a rough draft of the text to be translated, which is then corrected by professional translators. It can speed up their output by up to 80%. I know of one university that trains its students to use a TM tool known as TRADOS. They can then slot more easily into jobs as professional translators when they graduate. I often use Google Translate in the same way – but only with languages that I know reasonably well.