My next 4 blog posts will be about 4 mega trends that I think have one thing in common. They can all take two directions, heaven or hell. It is up to us what direction they will take. The fifth post will be a summing up and some conclusions. First something about my little experiment.
We are many who do not have English as our mother tongue. We must write in a foreign language, and although I feel quite at home with English it is indeed difficult to be equally nuanced in English as in my own language.
- Image via Wikipedia
There are many people who can not write or read English and for them the life on the internet is clearly limited. This is where machine translation comes in. When can we automatically translate a text with the such a quality that you can not see that it is translated? When can we automatically translate with the same quality as a good human translator? It will take a while that’s for sure. There is an excellent article in wikipedia about machine translation so I will not go deep into this field. Let us just note that it has evolved much slower than initially expected. There are many problems with automatic translation and it is clear that in order to make a good translation the translator must have some knowledge about the text’s context. That is something computers are not particularly good at.
There is a method that (relatively) recently took the field of machine translation one step further. The method is called “statistical machine translation“. The program automatically creates rules by going through large amounts of texts that are prepared in a special way. So far we have no such prepared texts in many languages, but it will come!
So far we have only talked about the written text. Don’t forget that to translate the spoken word is much more difficult.
Now for my little experiment
I will write this blog in Swedish and have it translated the to English with the Google translator.
I will correct the obvious errors that I can find but change as little as possible in the text. Every now and then I will write in English directly. Usually when I write english text I sometimes have to check some words meaning or spellings in dictionaries but the text is written in English.
Can anyone see through what is a machine translation and what is coming directly from me? Let us see!
For your information: This post is a combination. I first translated the text with the Google translator but after that I changed a fair amount of the text and also added more directly in English.
“When can we automatically translate a text with the such a quality that you can not see that it is translated?”
You blog post is interesting. Thank you.
I think that quality has two parts: fluency and accuracy. If the purpose of machine translation is to give information to people in their language, accuracy is more important than fluency. If a machine translation is correct, the fluency of the translation does not matter. For a small evaluation of the accuracy and the fluency of machine translation, see http://www.international-english.co.uk/mt-evaluation.html.
Interesting experiment. I got your point but in the same time I wonder if it is possible to be accurate without some degree of fluency and even more being fluent without being accurate. Just a short second of reflections…
I agree with you. Fluency and accuracy are related.
A translation can be fluent and not accurate.
A translation can be accurate and not fluent. If a translation has very low fluency, people cannot understand it. Therefore, it is not accurate.
Sorry that I was not clear in my previous message.