TM-Town Expert Translator Q&A

Adding Speech to Text to Your Workflow

Featuring Michael J.W. Beijer

Starts: November 19, 2015 Ends: November 21, 2015

Michael is a full-time professional translator with over 20 years of experience working in the European translation industry. He is well-versed in many translation tools and he uses speech to text technology to improve his productivity and make his life easier.

Browse All Q&A's
michaelbeijer

This Event Has Ended

Register for your free account so you don't miss the next expert Q&A event.

Questions: 9
Comments: 27

5

krasinskas

Saul Krasinskas

What software do you use for speech to text in your translation work?

3

michaelbeijer

Michael J.W. Beijer

Hi Saul,

I am currently using Dragon Professional Individual 14 (the latest version, which came out recently) on Windows 10. Dragon supplies the voice recognition functionality, but I also use a number of other tools (most of them open source), to add extremely powerful macro/command functionality to Dragon:

(1) Vocola (see: https://www.youtube.com/watch?v=d7BDTNNLQnk + https://www.youtube.com/watch?v=dgjWuHhD7zA + http://vocola.net/),
(2) Unimacro (see: http://qh.antenna.nl/unimacro/), and
(3) KnowBrainer 2015 (see e.g.: http://www.knowbrainer.com/forums/forum/categories.cfm?catid=3&entercat=y + http://www.knowbrainer.com/NewStore/pc/viewCategories.asp?idCategory=92).

I am also currently using Mark Lillibridge's amazing trick ("_select_and_say_test.py") to get Basic Text Control working in any Windows program/text field. For info on this, see our current discussion in the VoceCoders Yahoo list: https://groups.yahoo.com/neo/groups/VoiceCoder/conversations/topics/8434

1

michaelbeijer

Michael J.W. Beijer

PS: Oh yeah, and of course AutoHotkey (which is one of the most important reasons I could never switch to a Mac). Info @ http://ahkscript.org/

2

alinea-doc

Hans Lenting

Oh yeah, and of course AutoHotkey (which is one of the most important reasons I could never switch to a Mac).

Please stop teasing me!

1

michaelbeijer

Michael J.W. Beijer

Yes, that one was for you :joy:

1

hans

Hans van den Broek

MB: Thanks for your characteristically sourpuss take on things.

Well, you just shouldn't have said AutoHotkey (which is one of the most important reasons I could never switch to a Mac). That cannot be true. Unless you have no idea of Automator and AppleScript, of course.

By the way, Apple can come up with a lot of languages, and not all of a sudden, because text<>speech has been a part of OS long before OS X, and long before Nuance even existed. And as long as Nuance exists (and even when it still was L&H), rumour had it that Apple would buy it. They didn't. The most likely reason for that is that they were working on the technology themselves.

1

krasinskas

Saul Krasinskas

I'm using Typing Assistant (sumitsoft.com) and it's the reason of my not being able to switch to Linux:)

1

alinea-doc

Hans Lenting

With the risk of being off-topic: CafeTran's (Linux too) auto-completion feature is very good. It may well make your need for Typing Assistant disappear.

0

hans

Hans van den Broek

Now I don't want to be the party-pooper (you bet that's exactly what I want!) but:

*Dictation for OS X is excellent
*It's free (it comes with OS X)
*It covers far more languages than Dragon (all free, of course)
*Automator (plus AppleScript, Terminal, and other languages, all free) is both easier and more powerful than AHK

0

michaelbeijer

Michael J.W. Beijer

Thanks for your characteristically sourpuss take on things.

(1) I don't mind if OS X's built in Dictation app is good. In fact, I think that's great. I just don't have any experience with it, and know only a single translator who actually uses it in his daily translation work ("Tom from London"). From what I have read about it, I also suspect that its recognition engine is not as good as Dragon. However, I may be wrong, and there are also people who think that Apple licensed it from Nuance, which would mean it's exactly the same.

(2) It's nice that it's free, but it isn't of course. It comes at a price: the price of your Mac, and another price, but I won't go into that one ;)

(3) I do wonder about all those languages. I mean, how good can they all be? If it took Nuance years to create Dragon, how can anyone suddenly created hundreds of equally high quality engines? Nuance has a bunch of languages too, but only in its mobile apps. There is some talk that it will be adding more, also in connection with its upcoming "Dragon Anywhere", which will be available on mobile devices (iOS and Android; see: http://www.nuance.co.uk/for-business/by-product/dragon/dragon-anywhere/index.htmI + http://www.theverge.com/2015/8/18/9165425/nuance-dragon-anywhere-ios-android-dictation-app), and which can be connected to your local copy of Dragon Professional Individual 14.

(4) Automator is nice, but it's sadly not anywhere near as powerful as AHK. Also (and this is equally important to me), AHK has a really active and great community going on, whereas I doubt the same applies to the "Automator community", if such a thing even exists. I might be wrong though, as "I am not a Mac", as you would say. Just take a look at the AHK forums: https://autohotkey.com/boards/ (24,364 posts in the help section!!!!).

But AHK is just a tiny piece of the puzzle: what's really fun and exciting is adding Vocola, Unimacro, KnowBrainer and Dragonfly to the mix.

For a very interesting survey of ALL the speech recognition software available, both on Windows and Mac/Linux, see these two amazing blog posts: http://explosionduck.com/wp/introduction-to-voice-programming-part-one-dns-natlink/ + http://explosionduck.com/wp/introduction-to-voice-programming-part-two-open-source-speech-tools/

0

krasinskas

Saul Krasinskas

Michael, thanks a lot for your detailed response, I'm eager to check it all out later (busy right now), been thinking about using voice recognition for a long time, this might finally trigger it, thanks again

4

diasks2

Kevin Dias

How long did it take you to make the switch from typing to using speech to text in your daily life to the point where you "felt comfortable" with it? A few days, a few weeks, a few months?

1

michaelbeijer

Michael J.W. Beijer

Hmm, as far as I remember, it was a gradual thing, where initially I only used SR for writing emails and posts in mailing lists, etc., and maybe the occasional small translation job in my CAT tool, until, over time, I suddenly found that working on a computer without Dragon installed is quite distressing.

3

samuelholden

Samuel Sebastian Holden Bramah

Hi again Michael.

I was also wondering if you know of any good training courses specifically aimed at Translators wanting to kick off using Dragon and all these add-ons that you have mentioned... The learning curve seems quite steep and a bit daunting (and I am quite computer literate and not at all afraid of learning how to write these short scriptlets to make macros work).

I guess my main concerns are paying for the software and then not being able to get into the flow of using it, and also how much it is going to slow me down for the first month.

2

michaelbeijer

Michael J.W. Beijer

Hi Samuel,

I'm afraid that I don't know of any training courses aimed at helping translators get up to speed with speech recognition. If I had more free time, I might consider setting something up myself.

Currently, there are several very good places for information, and I highly recommend all three of these:

(1) Speech recognition for translators Yahoo Group: https://groups.yahoo.com/neo/groups/SR_for_translators/conversations/messages
(2) Speech recognition forum on Proz.com: http://www.proz.com/forum/speech_recognition-238.html
(3) The KnowBrainer forums: http://www.knowbrainer.com/forums/forum/index.cfm
(4) VoiceCoder Yahoo Group: https://groups.yahoo.com/neo/groups/VoiceCoder/conversations/messages (for the more technical among us)

Also note that the guys behind KnowBrainer have very graciously created a special subform (called "Third-Party Command Utilities - Vocola, Unimacro, VoicePower, Python... "), dedicated to open source SR software, etc.

Regarding having to learn to write and write scriptlets/macros, etc., one good thing about KnowBrainer 2015, for example, is that you don't have to write any scripts to get started, as the program actually ships with hundreds of them already. See: http://www.knowbrainer.com/NewStore/pc/viewCategories.asp?idCategory=92

Regarding your concerns about paying for something that you might not end up using, this is how I solved it (please don't tell anyone). I initially started using a cracked version of Dragon, which allowed me to use a fully functional professional version for as long as I felt was necessary to decide whether I wanted to buy it or not. Since I am a busy person, a 30-day trial simply won't cut it. I need to be able to use a program for several months, sometimes even up to a year (usually for many small periods, here and there, whenever I have some free time), before I really know whether it's for me or not. In the end, I purchased a licence for Dragon NaturallySpeaking Professional, and have recently (happily) upgraded to Dragon Professional Individual 14.

1

samuelholden

Samuel Sebastian Holden Bramah

Thanks! I will look into all that ;)

3

samuelholden

Samuel Sebastian Holden Bramah

Hi Michael.

I do a lot (around 80%) of my work in Trados Studio 2014.

Most of my work is technical manuals, scientific papers, medical records and heavy industry manuals, which means a lot of number crunching in Trados and a lot of copy overs to target segments, as well as re-wording of machine translated segments.

Can you tell me if you have any experience specifically with Trados and how you find it with that kind of translation?

1

michaelbeijer

Michael J.W. Beijer

Hi Samuel,

I'm afraid I have no experience with using Dragon in SDL Studio, although I have heard from various colleagues that Studio is supposed to be one of the less good programs for using with Dragon (as it doesn't allow Basic/Full Text Control). The best, currently, are apparently memoQ and Dejà-Vu. And, of course, Microsoft Word.

However, as I have mentioned elsewhere already, Mark Lillibridge (who is one of the main people working on Vocola 2 and NatLink.) (see: https://groups.yahoo.com/neo/groups/VoiceCoder/conversations/topics/8434) recently developed a workaround, allowing Basic Text Control in any text area on your computer, which would make working with Dragon in Studio a lot better (as this would allow you to not only dictate words in your target pane, but also select and/or correct them). If I have a moment, I will make a short screencast (and stick it on YouTube), testing Dragon with Studio 2014!

0

samuelholden

Samuel Sebastian Holden Bramah

Cool! That would be great to see!!

Thanks!

2

tomwalker

Thomas H Walker

Hi, Michael,
I have been interested in using SR to get an improved Source Text, when all I've been given is a really crappy PDF scan of a printed document. Some legal docs have initials or signatures scrawled over the body of the text, which further impedes getting a good clean ST using a document scanner with OCR. It often takes a long time to type in, or correct a scanned MSWord doc of, the ST. My thought is that I could read the text into an SR (usually I can see it better than a document scanner with OCR can), & save some time.
Have you used Dragon for this purpose? In order to be able to use the Dragon software in Spanish & in English (which I translate into), do I have to have 2 versions of the software? Or a special version? Or special add-ons? Appreciate any clarification.
Tom

1

michaelbeijer

Michael J.W. Beijer

Hi Thomas,

That sounds like a very good idea! In fact, my wife has done exactly this on several occasions, and I know that there are quite a few people out there who also use SR for data entry work, where they might, e.g., receive a pile of papers and dictate them into electronic form.

In fact, I even know someone who uses Dragon to help them transcribe hard to understand audio material. What they do is listen to the difficult to understand speaker (e.g. a YouTube video with c conference speaker) through headphones, while dictating what they hear into a Word document using Dragon. Sounds far-fetched perhaps, but it actually works!

2

diasks2

Kevin Dias

What are your thoughts around headset vs. stand alone mic? Also, does the quality of the microphone matter much, or will anything do?

3

michaelbeijer

Michael J.W. Beijer

Well, the headset vs. stand alone mic issue is a hotly debated topic, and in my experience there are a few people who would rather die than use a desktop mic, swearing that headsets are way & way better than any desktop mic could ever be (Jim Wardell, I'm looking at you)(see the recent Proz.com panel: "Leveraging voice recognition technology for greater efficiency panel" @ http://www.proz.com/virtual-conferences/668/program/10534 where we talked about this a bit). However, as you might already have guessed: I am a big desktop mic fan. I have never used a headset mic, and never will. I think that having to have one of those things strapped to your head would take away quite a lot of the magic of using SR, and make me feel like some kind of cyborg.

I am currently using a rather expensive ($279.00) desktop mic, the Speechware USB 3-in-1 TableMike (which I bought through KnowBrainer; http://www.knowbrainer.com/NewStore/pc/viewPrd.asp?idproduct=402), but have gotten equally good results using a cheap Logitech desktop mic! However, not all desktop mics perform the same, so it is very well possible that yours will perform abysmally. There is only one way to find out. I actually did quite extensive testing, between my expensive TableMike and my very inexpensive Logitech desktop mic, and could hardly tell the difference, if at all. This is the Logitech one I tested: http://www.amazon.co.uk/gp/product/B00009EHJV/ref=oh_o00_s00_i00_details

Oh yeah, my TableMike also has an excellent noise cancelling system and I can dictate while listening to background music.

My wife is currently using the Logitech one with Dragon Dictate for Mac, and getting great results.

Image of Logitech desktop mic

The speech recognition software you are using will also make a difference here. For example, newer versions of Dragon Professional have something called a "Far Field" acoustic model, which is designed especially for microphones that are farther away, i.e. built-in laptop mics and desktop mics. Older versions of Dragon don't have this, so will perform less well with desktop mics. Sadly, however, the Far Field model is only available for a few languages (e.g., for some reason, it is available for US English, but not UK English), but I expect Nuance will change this in the future.

Dragon NaturallySpeaking13 includes support for end-users that use Dragon with a
laptop that has a built-in microphone. This type of microphone is considered to be
‘far’ from the user’s mouth when they are performing dictation.

The user may use the Far Field BestMatch IV acoustic model or the Far Field
BestMatch V acoustic model when they create a User Profile. Both of these models
are available in the Speech acoustic model drop-down list on the Choose
speech recognition options screen.

(Dragon info pdf @ http://goo.gl/Gicwv7)

0

diasks2

Kevin Dias

Thanks! Good to know. It would drive me nuts to wear a headset mic all day.

2

hans

Hans van den Broek

I also need that information badly. I now use the built-in microphone, and I probably need a noise cancelling one (aircon or fan, or both). And it'll have to be a cheap one because I'm not sure at all I want to use speech recognition.

2

hans

Hans van den Broek

I tried dictation more than a decade ago, and voted against it for several reasons. I'm prepared to try again - I actually did try again very recently - but I encountered a "new" problem: The language for the commands. Everything seems to be fine when you dictate in English, but if you dictate in another language, the commands will be in that language. If available. Since I write in English, Dutch, and Bahasa Indonesia on a daily basis, and since I never changed the language of the OS nor of individual applications, I'm deep in it. I know the commands in English of course, I can guess them for Dutch, but I'd have to learn all of them for Indonesian.
Bahasa Indonesia
Even worse, if I use CafeTran, and want to dictate my translation in Dutch, how can I handle those commands? CafeTran hasn't been localised into Dutch, so the commands will have to be in English. Would that be a problem? (I haven't tried. What I did try is add a marked down pic in this forum, see what happens)

1

michaelbeijer

Michael J.W. Beijer

How about some Java syntax highlighting Markdown:


Hmm, I know very little about juggling multiple languages, and am one of the lucky ones, as all I need is English.

I suppose that if I did want to dictate into, e.g., German, using the German version of Dragon, all my Vocola/KnowBrainer commands would still be in English (since I defined/created them myself), so I wouldn't face the same problem as you.

Is there no way on the Mac to, say, dictate in Dutch, but keep your OS in English, and all your Dictation commands in English?

PS: Markdown = cool!

0

hans

Hans van den Broek

My OS is in English, and so are my apps. Always have been, and that's part of the problem (I think). All standard commands are in the dictation language, and I don't know them, especially not for Indonesian. I can change them, but that's a lot of work. In CafeTran, it gets really complicated: OS in English, dictation language Dutch, standard commands Dutch, CafeTran-specific commands English. O well, I should just try. Thanks, I up-voted you (but don't you think for a moment I'll make a habit of it). Now the microphone question, please.

1

diasks2

Kevin Dias

This question isn't necessarily just for Michael, anyone can feel free to chime in. I'm thinking of adding an optional field to the TM-Town profiles related to whether one uses speech to text software. It would show on the left hand side of the 'About Me' as an icon (maybe under CAT tools). What are your thoughts?

My feeling is that most clients won't care...but one purpose of TM-Town is also to help translators connect with peers in the industry, and this may be an interesting data point in that respect.

0

michaelbeijer

Michael J.W. Beijer

Sounds like a good idea to meet Kevin!

1

diasks2

Kevin Dias

We often hear of the benefits of using speech to text - what have you found to be the biggest negative or downside?

3

michaelbeijer

Michael J.W. Beijer

One downside of using speech recognition (SR) is that it can sometimes be harder to move words or phrases around or correct things by voice than with your keyboard or mouse. Having said that, this usually isn't such a problem as I like to combine speech recognition with keyboard/mouse use, rather than just try to do everything by voice alone, or keyboard/mouse alone. If you ask me, this is where the real benefits of SR start to appear: once you learn how to juggle the various input methods smoothly and enter flow.

For example, I am writing this reply using a combination of Dragon and my keyboard and mouse. For example, I might start dictating a sentence and then edit it using my keyboard and mouse. or I might just dictate the entire sentence and do any editing by voice alone. It will really depend on the specific thing that I am trying to do at any given moment. The better you get at using Dragon's various correction and editing features, the faster you will get at using voice alone. However, some things remain faster using a combination method.

Another potential downside involves lack of support for a particular program on your computer. Although this is an entire topic in itself, I will quickly try to explain what I mean. In Dragon (I know that there are other SR solutions out there, but 90% of people reading this will probably be using Nuance's Dragon, in one form or another), there are various levels of so-called "text control" (Dragon dictation support in a text field) that may be available in a particular program, the highest level being "Full Text Control" (Dragon allows you to select words you just dictated, and then edit them using various commands such as "correct that" and "spell that"), which is followed by "Basic Text Control". Sadly, Dragon only provides Full Text Control, or even Basic Text Control, in a very limited number of Windows and Mac programs, primarily limited to very standard ones such as MS Word, Notepad, and a few others. Recently, however, someone in the open source community managed to devise an ingenious workaround (involving Python and the NatLink project), but this is a topic in its own right and I won't go into it here.

Another downside, which isn't really a downside, is that when starting with SR, you might have to rethink how you go about translating or doing things. Some people find the change quite difficult to deal with, and end up giving up. However, if you are like me, you will soon enough discover that it is more than worth it, and that these changes to your workflows are all positive. I have noticed, for example, that dictating my translations has greatly improved their flow and readability, which is yet another topic I would like to expand on elsewhere.

2

reedjames

Reed James

Hi Michael,

Another important drawback to using SR are the misrecognitions that invariably end up in your translations unless you are very careful and thorough in proofreading them. Have you ever let one of these misrecognitions out "into the wild"?. How do you compensate for this? Though a Dragon user + KnowBrainer + VoiceComputer, I am many times reluctant to dictate my translations for this very reason.

Reed

1

michaelbeijer

Michael J.W. Beijer

Hi Reed,

Well, if I let one "into the wild", no one mentioned it ;)

However, I think the same thing can be said about translation using any input method (SR, keyboard, CAT tool, pencil): mistakes can (and therefore will) be made. I think I actually make less typos dictating than typing, but that's just me. After a while you also get pretty good at spotting the special types of errors made when dictating, which are different in kind than those made when e.g. typing.

I'm also never reluctant to dictate my translations, as I always carefully proofread every word I write before sending my translation off to the client.

0

diasks2

Kevin Dias

Thanks Michael! I'm curious, for example in the answer you just dictated above - can you do things like bolding words (i.e. in this example wrapping the word in double asterisks) with dictation, or is that something you use the keyboard for? Is this something that you would need to program a specific shortcut for?

2

michaelbeijer

Michael J.W. Beijer

Sure, that's easy! There are various ways that I could do this.

I could create a little command in Vocola to do it. Creating new commands in Vocola is extremely simple. All you have to do is say "edit commands" (all commands for the program you are in) or edit "global commands" (all commands that apply globally on your computer) and a text file containing all of your Vocola commands will open in the text editor of your choice where you can create the new command. In Vocola, each command is written on its own line. For example, here is a simple command to search for a selection in Xbench:

search Xbench = Keys.SendInput({Ctrl+Alt+x});

However, in this particular case, because I already have various Autohotkey commands that do similar things, it would probably be easier to just invoke these AutoHotkey commands via Vocola. It is currently also possible to use Autohotkey code directly in Vocola commands (!). See e.g.: http://qh.antenna.nl/unimacro/features/autohotkey/index.html

Anyway, back to the question. So let's say I wanted to make selections bold using Markdown by voice. I could then trigger the following AHK command from Vocola:

-----------------------------
;command to surround selection with Markdown **double asterisks** to make it bold

#!b::                                               ; keyboard shortcut = win+alt+b
Send ^c                                           ; copies selection to clipboard
Sleep 100                                        ; wait a tiny bit of time
clipboard = **%clipboard%**            ; surround the clipboard content with **
Send ^v                                           ; paste edited clipboard
return
---------------------------

To trigger this AHK script via voice, I would say "edit commands" to call up my list of Vocola commands, and add this line to the file:

markdown bold = Keys.SendInput({Win+Alt+b}); ### surround selection with Markdown **double asterisks** to make it bold

The above Vocola code means: when I say "markdown bold", Win+Alt+b gets automatically pressed, triggering my AHK script.

Although this might sound a little confusing or complicated, I assure you that once you have mastered it, it is extremely quick and easy to do. Also, people might think that using these AHK and Vocola commands might be slow to execute. However, this is not at all the case: all of these commands execute lightning fast and both Vocola and AHK are extremely light on resources. Dragon, on the other hand, can be a bit heavy (on your CPU and especially RAM), but this is another topic in its own right, deserving proper treatment elsewhere.


Are you a professional translator?

Start putting your work to work for you.

Join TM-Town Today!