Some of you may know that I started my professional life as a translator. I studied French and English at Saarbrücken University. I had a few freelance translation assignments before joining Trados in 1994 as the first dedicated training and support engineer. Yes that is a long time ago…!
The reason I’m starting with this background is that even today, I am still trying to do the odd translation assignment whenever my schedule allows for this. This enables me to ‘touch’ SDL Trados Studio and MultiTerm in real-life scenarios that go beyond the sample photo printer document that we at SDL like using so much for product demos in all shapes and sizes ;).
Typically I translate two types of documents – one, software string files either in SDL Passolo or SDL Trados Studio, and two, PowerPoint slide decks for events such as Tekom, where the German language still prevails and has not yet been fully overtaken by the virtually omnipresent English language spoken at conferences. So it can happen (quite frequently, actually), that I need to translate my own English slide decks into German. As much as I try to make the time for my translation activities, there are many other things to attend to as a product manager, so I often leave translation tasks to the last minute.
This is where I had a kind of light bulb moment the last time I left things to the very day before Tekom started. Two fairly large slide decks were waiting for me. I knew I had to try and get through the translation fast as there was literally no time left. At the same time, PowerPoint presentations are not known to be very repetitive so I could not hope for translation memory or even termbase matches to speed me up too much. So I thought this might be an interesting time to try out what was to become one of the key features in our new Studio 2015 release: using Machine Translation as a source for AutoSuggest predictive typing.
I will admit that I was not too hopeful at first since in my mind marketing language and Machine Translation (MT) are not exactly a ‘perfect match’. And indeed, when you work with MT in the traditional way where you post-edit draft translations, it can be frustrating to work with translations that are often not good enough and have to be changed quite drastically, or even started from scratch, to reach acceptable quality levels.
However, with AutoSuggest 2.0 drawing from Machine Translation, I thought it would be good to give MT a fresh start, as in this case you do not post-edit translations, but rather pick and choose the bits and pieces from MT that fit into the translation as you work. So in a way this puts me as a translator in control of MT, rather than the other way round.
As I started to work on the presentation, I was pretty amazed at those ‘bits and pieces’ that MT kept suggesting to me to help me speed up my work. I found that when you don’t work with MT at the entire sentence level, but rather at a fragment level, it’s like a constant source of inspiration. You are free to ignore the bad suggestions (which are of course still there) and at the same time will constantly get good suggestions that fit well into the current text flow.
To cut a long story short, once I got used to the way MT was suggesting fragments to be as I typed, I was literally whizzing through the translation and was able to finish it on time. It’s great to see that we are now able to offer this feature to all our users and I hope you will enjoy this way of working with MT as much as I did. I feel it is a long overdue innovation to bring MT to translation processes in a very attractive way.
Please join me again tomorrow, where I will be running through another new exciting feature called Retrofit.