ITI Bulletin article: Taming the Dragon
Ellen Worrell reports from the ITI's Dictation for Translators workshop that focused on using automatic speech recognition software to increase productivity and improve translators' health
I suffer from what I have termed netballer's finger – I've lost count of the number of fractures I've had and ligaments I've damaged over the years, and my fingers can suffer after a long day's typing. So when I heard of other translators using Dragon to dictate their translations, I thought it could be just the thing my poor fingers were after.
I took the plunge and invested in Dragon Naturally Speaking about six months ago and have slowly gotten used to dictating my emails and straightforward Word documents. Unfortunately, I'm yet to fall in love with Dragon when translating. I think that is partly out of apprehension and partly out of frustration:
How do I handle tags?
How can I spell German, French, Spanish names?
Can I confirm segments using my voice rather than the keyboard?
When John Moran's dictation workshop was announced, I thought it sounded like the perfect opportunity to ask these burning questions.
John – a German to English translator himself – is very interested in studying and increasing translator productivity. At the beginning of the workshop, John explained that machine translation can help increase productivity but only in certain situations. In contrast, using speech recognition software such as Dragon Naturally Speaking can help increase productivity even further.
Research suggests average productivity gains from using a CAT tool are about 30%, but these tools may have decreased productivity overall due to their general incompatibility with speech recognition software. Most translators who dictate using Dragon suggested their productivity gains start at around 30%. The ITI and CIOL conducted a survey in 2011 and, of the 1750 respondents, 55% of people said they used CAT tools but only about 10% used automatic speech recognition. John is very keen to find out why so few translators seem to use automatic speech recognition when it not only increases productivity but also offers a wide range of health benefits.
A study conducted by Ehrensberger-Dow et al in 2015 found that "some of the most frequent complaints (pain in the neck/shoulder, burning eyes, visual fatigue, pain in arms or hands, back pain) would seem to be directly related to intensive screen work and activities associated with inputting text at a computer". By dictating your translation, you are not tied to the keyboard, and – more importantly – you can keep working even if you are in pain. Since translators hit an average of 20,000 keystrokes a day, learning to translate using Dragon may well save my fingers in the long run.
Before we were all let loose on our Dragons, John ran through a few brief pointers: Dragon needs a minimum of 8 GB of RAM and having a solid state hard drive does not hurt either. A high-quality microphone that minimises background noise is also a worthwhile investment. Dragon is highly accurate in English and German, and is gaining popularity in other languages such as French, Spanish and Danish.
John had asked us all to bring texts that we found easy to translate so that we could practice in the second half of the workshop. As we translators are a varied bunch, there was a whole gamut of texts – from website copy for patio heaters right through to gazelles! John uses Dragon when he is translating and said that once he gets into the rhythm of dictating a translation, it can feel a bit like interpreting which then urges you to keep going at a natural pace. He did admit, however, that Dragon is not great for texts where there are numerous tags or that require a lot of research. One of the other attendees suggested that for research-intensive texts, it could be beneficial to spend a couple of hours at the beginning of the project researching terms and entering them in a termbase so that you can then focus on dictating your translation rather than forever interrupting your dictation.
We then split into different rooms to have a go at taming our Dragons. In pairs, we tried to master the various commands and what to do if we want Dragon to type one of these commands. Many Dragon aficionados will know that you can simply turn the microphone off by uttering "go to sleep", but how can you dictate that phrase without turning the microphone off? The answer is simple: hold down Shift while dictating. We also learnt that to confirm segments, you can say "type control and enter". We explored ways to edit vocabulary and even how to use dual-language dictation to minimise homophones such as "I" and "eye" (tip: train Dragon to recognise the source-language word but type the target-language word). Finally, John showed us how to train Dragon using material we have previously translated, helping the software to recognise our writing style, thereby improving accuracy in the long-run.
It was a very interesting workshop and has given me the confidence to persevere with Dragon while translating. My first step: to dictate this article!