text the phrase 'I would like an apple' appears 90 times, then that would seem to be a correct translation with 90 per cent confidence. Maybe one of the original lines

said 'I would like an apple and a pear' and another said 'I would like an apple and a banana'. Since the pear and banana are only mentioned once each in the translations, the computer knows to leave them out. The computer doesn't have to understand anything about language to do this. Given enough sample text, the computer could translate Shakespeare into Dothraki, the language invented by the makers of Game of Thrones. Google Translate was developed under the leadership of computer scientist Franz Och and originally used documents created by the United Nations. The UN has a need to create a lot of written material accurately in a number of languages, and all of this material is freely available, which made it a good source to use. (We should point out that Google did not invent statistical translation.) 2.1 Try the exercises suggested in the case - how did Google Translate do? [5] 2.2 Could this technology be used by a global firm or is the need for absolute accuracy so large that a human translator would always be needed? [5] 2.3 What are some of the problems with using Google Translate to translate a web page? [5] 2.4. What would the computer need in order to translate Shakespeare into Dothraki? [5]/nQuestion Two Translating Shakespeare into Dothraki [20 Marks] Many people using Google Translate for the first time are amazed by its abilities. If you haven't tried it, visit your local site (it's https://translate.google.co.uk in the UK), enter some text, choose a language and click translate. If you actually speak the language you translated into, how did it do? Does the translation make sense and does it say what you meant it to say? If you don't speak that language, copy the translated text and translate it back into the first language - it probably won't be exactly the same, but does it still mean what you said in the first place? Try translating the text through several other languages - English to German to French to Chinese and then back to English - what does it say now? This game is actually popular fodder for YouTube videos - YouTubers translate the lyrics of a popular song (usually a song in English) into another language, then back into English, and record themselves singing exactly what they ended up with. Clearly, however, this technology has more serious applications. So how does it work? It relies on the one thing that Google is enormously good at - processing lots and lots of data, very very quickly. Sorry to disappoint you if you thought the computer could actually understand what you are saying, but in fact it's just doing some statistics. There are many books that have been translated by a human from one language into another, say Victor Hugo's Notre Dame de Paris from French to English. To translate a French phrase, all the computer has to do is find the same phrase in one of these books, then find the corresponding page and line in the English translation and return that to the user. But how does the computer know that the page and line are the exact bits that translate the exact phrase the user wrote? That's where the statistics and data processing come in. The computer doesn't just search for the phrase in one book, it looks in say 100 books (the actual figure Google uses is difficult to track down) and finds 100 possible translations. It then chooses the phrase that appears the most in the sample of 100. For instance, if the user writes 'Je voudrais une pomme pour mon petit déjeuner avec mamie', this line might not appear in any book, but the phrase 'Je voudrais une pomme' probably does. If the computer finds 100 books where this line appears, and in the translated book at around about the right point in the