Using Dragon Naturally Speaking in translations

translation_articles_icon

ProZ.com Translation Article Knowledgebase

Articles about translation and interpreting
Article Categories
Search Articles


Advanced Search
About the Articles Knowledgebase
ProZ.com has created this section with the goals of:

Further enabling knowledge sharing among professionals
Providing resources for the education of clients and translators
Offering an additional channel for promotion of ProZ.com members (as authors)

We invite your participation and feedback concerning this new resource.

More info and discussion >

Article Options
Your Favorite Articles
Recommended Articles
  1. ProZ.com overview and action plan (#1 of 8): Sourcing (ie. jobs / directory)
  2. Réalité de la traduction automatique en 2014
  3. Getting the most out of ProZ.com: A guide for translators and interpreters
  4. Does Juliet's Rose, by Any Other Name, Smell as Sweet?
  5. The difference between editing and proofreading
No recommended articles found.

 »  Articles Overview  »  Technology  »  CAT Tools  »  Using Dragon Naturally Speaking in translations

Using Dragon Naturally Speaking in translations

By Janis Abens | Published  02/28/2007 | CAT Tools | Recommendation:RateSecARateSecARateSecARateSecARateSecI
Contact the author
Quicklink: http://esl.proz.com/doc/1178
Author:
Janis Abens
Letonia
sueco al inglés translator
 

See this author's ProZ.com profile
Speech recognition has been around for some time now, yet relatively few "normal" computer users have adopted this technology. Perhaps they tried it out in the early years, and were dissatisfied with the speed and/or accuracy Maybe the need to train the program and enunciate words clearly was a barrier. Often, people who type quickly abd accurately think that the increase in speed is either nonexistent or not worth the effort. Many users are only vaguely aware that the technology exists or think it is only for the handicapped.

My experience has been purely positive. After developing pains and aches slaving over a huge job, I discovered DNS way back in v.5. The program took abot 20 minutes to get going, and I was immediately amazed to see I could speak in a normal tone and at normal speed with very few errors, just watching the text appear on the screen in front of me as I spoke. I was even more amazed at the huge vocabulary, including all kinds of scientific, medical and technical terminology.

Obviously, not all types of texts are suitable for dictation, such as technical manuals etc. Also, the entire point of CAT is to not have to create every word anew, wheteher typing or talking. So, in reality, the way I use DNS/DVX varies between different jobs.

Speech recognition, however, is not solely for dictation. Specific voice commands can be assigned to keystrokes, combinations and even complex macros. Thusly, I assigned personalized commands to all the functions I use in my CAT tool, such as creating a new project, changing options like fuzzy match settings, navigating within and between segments, joining and splitting segments, saving segments to desired databases and so on.

So, even though I still type alot, whenever I plan on working for extended periods, I always use DNS. Even though it is still faster to press CTRL-A than to say "assemble", ther comes a time after X hours of drudgery when I can lean back in my chair or stand up and stretch, and continue working more or less unabated. This certainly comes in handy when the shoulders and wrists start to tell you its time for a break, but the deadline is still looming.

So for texts with tables, lists, etc. and many repetitions there is no need to abandon the CAT functions, and I use DNS mainly to control the computer, sporadically saying a sentence or two. Especially with long and convoluted sentences, the CAT result is often essentially correct but the sentence clauses are mixed up and need rearrqanging. Instead of click/drag (awful) or CTRL-arrow selection, shifting around etc. it can be much faster to just say the correct sentence.

OTOH, I often get a text in paper or image format. Sometimes it is worth the effort to scan, OCR, correct and proceed with the digital file. Often, however, it is much faster to simply translate on-the-fly, with my complete focus on the paper, not even looking at the computer. I say without exaggeration that I cannot speak fast enough to confuse or overload DNS. After learning not to slur and stutter, the program makes virtually no mistakes. Even homonyms are correctly recognized by the context if the proper training is done. (Letting DNS check your docs for samples of the context, or teaching it as you go.)

Is there a downside? Well, since DNS does not make spelling errors, a spell check will not reveal mistakes, and one will run into like-sounding alternatives to what you actually meant now and then. So proofreading is still of the essence.

In summary, there is no right and wrong way to incorporate speech recognition into a translation workflow. Just as people combine mouse operations and keyboard shortcuts in various fashions, the spoken word not only adds a third dimension to controlling the machine but can also be correctly reproduced as text on the screen


Leveraging the power of voice recognition and computer - assisted translation to optimize the entire process from workflow design to translation and especially PC/ network/ office management has given me an enormous boost in productivity. The learning curve is minimal and the improvement in the quality of life is surprisingly great! There is certainly no going back...

http://www.abens.org




Copyright © ProZ.com, 1999-2024. All rights reserved.
Comments on this article

Knowledgebase Contributions Related to this Article
  • No contributions found.
     
Want to contribute to the article knowledgebase? Join ProZ.com.


Articles are copyright © ProZ.com, 1999-2024, except where otherwise indicated. All rights reserved.
Content may not be republished without the consent of ProZ.com.