by Людвиг Главати (Ludvig Glavati)
"He found that replacing his typewriter with a Dictaphone not only increased his output but also eliminated the strain of typing each manuscript."
Erle Stanley Gardner by W. F. Nolan
If to compare translation and interpreting from pure technical point of view, the main difference is speed. Speech makes about 3000 words per hour and it is natural process. Translators type much slower, switching attention alternately between reading, typing, checking in dictionaries and so on.
Of course, when you dictate a document your output can't be a high quality translation, it needs corrections and proofreading.
From my experience, I estimate that text obtained by dictation is about 70-80% text job done. I call it "raw translation".
Another big advantage lies in fact, that speaking aloud of draft translation rapidly "downloads" the text to your head, which makes the following corrections much more easy.
In the process of dictation your mind quickly runs over different variants of phrases. You can also dictate those different variants and later select the best ones in the obtained text.
For years, I transcribed my dictation files by typists, and sometimes by myself. As there was no working software for Russian speech recognition.
The breakthrough was made by Google, though, only as voice input on websites. In 2013 I found a website, where you can convert your voice into text documents - Speechpad.pw (Russian version - Speechpad.ru). Since then, I use it for all my translations. Of course, it makes more mistakes then a typist, but still works quite good for me, and it is quite free and always available at good internet connection.
Apart from Russian, sometimes I input simple English and Ukrainian text (generally, when I help my children with their homework).
If you want your dictation to be easily convertible to Translation memory, the source sentences should be in strict order, so that later you could easily to align them with dictated translation. For this purpose I clean the source text from formatting elements, and also remove repetitions non-translatable segments with obtaining a source "core" for dictation (see my article "Alternative Translation Approach: Labor division").
Another issue is that dictation of bare text can become rather boring and demanding task.
Once, I borrowed the instrument used by TV new announcers. They dictate their news reading it from running text line - "marquee" - placed just above the camera, so it seems as if they look at viewers while speaking.
I use my Marquee tool, which converts the text to HTML-file with running line. Under the line there are Speed control buttons allowing to adjust it to your dictation tempo. Running text line holds attention, in the same way as a person you talk to, so, dictation is made easily, almost with no strain.
As I get a new translation job, the first thing I do - use my "TranslatorsFactory" toolkit for Word in order to prepare a one-column table with source text cleaned from formatting elements, repetitions, and non-translatables. Each cell of the table contains one source segment.
Then, I generate a marquee - HTML-file, which I run either on PC, or laptop, or android tablet - along with audio-recorder for dictation.
When dictation is completed, I convert the audio-record to text, insert it in the second column of the table, align it with source segments, and correct.
Finally, I convert the table to Translation memory (TMX-file).
After applying the TM to the source document, the rest is to correct it using a CAT tool.
Another quote from Erle Stanley Gardner's biography:
"In 1932 (!) Gardner began to dictate his stories on wax cylinders, turning them over to his secretary for transcriptions. ... Called "the Henry Ford of detective fiction" Gardner had typed 66,000 words a week".
This means approx. 5-7 hour dictation every day. And it's about year 1932, wax cylinders, and an author recognized to be classic of genre!
Someone could object that translation is quite different from writing.
Then here is another story.
Some time ago, an interesting post appeared at Proz.com's forum of Don Hank from the USA, who has been dictating translation (full time) for over 40 years, sending dictation files to his typist. Don asked "Is TRADOS worth it if I dictate translation?". Of course, there were many good answers. In the end he summarized:
"Thanks for the great answers. ... I guess I will not be buying any spiffy new tools for a while. My typist would starve anyway if I started typing. By the way, I have always liked the dictating approach because you never have to take your eyes off the page. AND the typist also creates all those tables and inserts the figures, jobs I don't like."
Hope, this man makes not less and not worse translation than many of those who use "spiffy new tools".
Just a while ago Machine translation was taboo for Language Service Providers. Today, increasingly frequently, quite solid companies use MT for localizing their websites and paperwork. And quite solid agencies offer proofreading of machine translation. Especially interesting, some of those jobs are corrections of mixed MT fragments, like eBay ads, which are used for development of Google translation engine utilizing Statistical Machine Translation technology. This is the "cutting edge". When performed those jobs I felt like an "appendix" to artificial intellect. Though, suppose it is an inevitable trend. At least for technical translation. (In fact, the vast majority of translations are technical.)
Dictation approach seems to be an undeveloped land. Though promising.
What is the real value of dictation - it remains Human language, while it's speed is comparable to Machine translation.
Людвиг Главати (Ludvig Glavati) is a translator and interpreter from Ukraine.