Accuracy Improving and Timesaving Transcription Tips

  • 2
  • Problem
  • Updated 1 year ago
  • (Edited)
I've seen a few questions on transcription lately, and since transcription and captioning is my specialty, I thought I would chime in and add some snips of insight I hope might be helpful...

Especially if you are in education and make a lot of ADA (or your country's accessibility guidelines) compliant videos without a script, you are probably spending way too much of your valuable Camtasia time transcribing your videos for captioning. Camtasia really won't mind if you check out a few tips and tricks, and maybe even try a far more efficient, albeit wacky looking, transcription method outside of her interface, that will ultimately free up lots more fun Camtasia screencast time.

Dictating, I use Dragon 15 or Dragon 6 for Mac and a Yeti Microphone on a boom... I generally get upwards of 95% accuracy.
Transcription (my voice) of a good sound quality .wav file averages about 85 - 92% accuracy.
Windows speech recognition on live dictation is nowhere close to dragon 15, and even less accurate when transcribing. It's fine for the occasional leisurely transcription, but if you have a lot to transcribe, don't bother with Windows speech recognition unless you have no other option.

If your video includes high quality audio (garbage in, garbage out is gospel when it comes to audio), your best solution might well be as simple as uploading to YouTube. The auto transcription has improved tremendously, and it's going to provide a caption file as well. I did a test recently on a 4 minute video with very good audio - 94% accuracy. Not bad!

A couple of tips...
Accuracy decreases as audio length increases on most computers with any transcription program. If you have a big file, make a copy that you can chunk into smaller pieces for transcription purposes.
Reboot your computer before you start. You'll be surprised how much that alone can improve accuracy.

I do a lot of transcribing for captioning. Since dictation is always more accurate than transcription, I've got a crazy little method that saves me ridiculous amounts of time.

I use 2 computers.
On one computer, I have Express Scribe Transciption Software (you'll need the paid version for video) and I plug in a USB Transcribing Foot Pedal (mine is Infinity) Videos and/or audio files To transcribe go here. I plug my earbuds into this computer.
On the second computer, I have Dragon Naturally Speaking and Microsoft Word. I have an Andrea NC-181 VM USB High Fidelity Monaural (one ear ) Headset microphone set up.

I run the video through Express Scribe, controlling play, stop, rewind and fast forward with the foot pedal. I listen with the earbud
I dictate what I hear into Word on the other computer, using Dragon, via the monoheadset. I don't listen and stop the video to dictate, I dictate while I am listening until I can't keep up, then I stop, back the video/audio up just a tad to where it got ahead of me, and start again. If you slow down the speed of the audio to a comfortable listening pace, (a feature in Express Scribe) you can pretty much dictate in just about real time. Very few people can type as fast as they can talk.If you enable punctuation in Dragon, on the newest version of Word, you'll get a ton of help with punctuation automatically. This method is efficient and SO MUCH FUN!

If you spend a lot of time transcribing, this method could be your new best friend. It will pay for itself almost immediately with time saved, and have you back in Camtasia screencasting In no time. I use MovieCaptioner(Mac and PC less than 100.00) and SubTitle Edit (PC free) to create captions - with my transcript or from scratch. Or, if Im uploading the video to YouTube, I upload my transcript and use YouTubes very efficient sync option to create captions/caption file.
Photo of susannemistric

susannemistric

  • 100 Posts
  • 76 Reply Likes

Posted 1 year ago

  • 2
Photo of Joe Morgan

Joe Morgan

  • 5650 Posts
  • 2929 Reply Likes
I agree, I use and endorse Dragon 15.
 Plus, an Andrea microphone NC-185 VM USB High Fidelity Stereo USB Computer Headset with Noise Canceling Microphone . Dragon gave it 5 stars and it's noise canceling feature is really good.

I disagree with you that Windows speech recognition is fine for the occasional leisurely transcription. Windows speech to text is horrendous for transcribing and always has been.  Perhaps your response is colored by the fact you are a re-seller of Camtasia.

Regards,Joe
Photo of Joe Morgan

Joe Morgan

  • 5650 Posts
  • 2929 Reply Likes
I've been down the Windows rabbit hole. I tried very hard to make Windows speech recognition a viable option for me.
I was to cheap to invest in Dragon Naturally Speaking at the time.

I invested a lot of time and effort, I trained the program individual words,  the ones the program wouldn't recognize on it's own.
Read story's to it from the internet. "Like older versions of Dragon encouraged you to do". The accuracy never improved much at all.

I have a Pragmatic approach to this. You can pick up Dragon 13 from Amazon on the cheap these days.



If your aware going in just how lousy Windows is, you might not waste your time the way I did.
Microsoft hasn't improved the program much over time. It appears to be the same piece of  junk it always has been.
My screen shot in my previous post below shows a 40% error rate over 49 spoken words. That's ridiculous.

I understand that some people may have no other options. 

 I hope you don't think I'm trying to give you a hard time, because that's not my intention. Not at ALL!

My intention is save others the aggravation of getting nowhere with Windows.

When I purchased Dragon 13. It wouldn't import my version 12 voice profile. So I had to start all over from scratch.
Out of the box, it still recognized nearly every word I spoke perfectly. Accuracy must have been up around 97% or  98%.

Windows with it's 40% error rate. I ran the standard voice training. Plus, I even did some additional voice training before running that test. That's truly unworkable results.  I'm running the latest version of Windows 10.

Regards,Joe
  
Photo of susannemistric

susannemistric

  • 100 Posts
  • 76 Reply Likes
I can't speak about Windows 10 speech recognition. Now, THERE is something I have avoided at all costs. None of my workflow software runs better on Windows 10, so I've managed to avoid it thus far. I always favor MACS, and now that Dragon and Camtasia have very nice versions for MAC, I can do much of my work on my iMac. The newest MAC OS has AWESOME and very easy to use extended native speech to text accuracy, by the way. Limited uses at this point, but typical Apple quality that will continue to expand and improve, I am certain
(Edited)
Photo of tony.lima

tony.lima

  • 35 Posts
  • 4 Reply Likes
We use Apple products almost exclusively.  Got one Win 7 machine here that we mainly fire up to do updates.  There are still a few bits of software that don't run on the Mac.  And, yes, I have VMware Fusion.
Photo of susannemistric

susannemistric

  • 100 Posts
  • 76 Reply Likes
I just can't bring myself to run Windows on my MACS...So I keep a PC for some tutorials and programs.
Photo of tony.lima

tony.lima

  • 35 Posts
  • 4 Reply Likes
Hey, I still have a Win XP notebook here.  Haven't powered it up for months.  You might like my tech blog: http://TonyLimaAssociates.com
Photo of docsope

docsope

  • 69 Posts
  • 33 Reply Likes
Really informative, thanks susannemistric  
Photo of Glenn Hoeppner

Glenn Hoeppner, Employee

  • 1754 Posts
  • 286 Reply Likes
This is great information. Thanks for taking the time to share!
Photo of Joe Morgan

Joe Morgan

  • 5650 Posts
  • 2929 Reply Likes
Yes susannemistric,

The information you provided is very helpful and I should have said that in my first post.

However, I stand by my assessment that Windows speech to text is horrendous and should probably be avoided like a grizzly bear.
 To be fair I tested a very short "In Editor" Voice Narration using the latest version of Camtasia 9.

The inaccuracy came in at 40%.
To me, proofreading and correcting it would take more time than doing it from scratch.

   This isn't TechSmiths fault, it's all on Windows.

 Regards,Joe
Photo of alixreed.abr

alixreed.abr

  • 23 Posts
  • 0 Reply Likes
Thanks so much for taking the time to share with us!
Photo of tony.lima

tony.lima

  • 35 Posts
  • 4 Reply Likes
Susanne, many, many thanks for the excellent tips.  Naturally I have a question.  Have you tried iPhone dictating?  I've found that dictating straight into the notepad app does a pretty good job.  (I can't use dictation to transcription because I'm working in economics.  Transcription vocabulary doesn't handle "heteroscedasticity" very well.)
Photo of susannemistric

susannemistric

  • 100 Posts
  • 76 Reply Likes
If you are using Dragon, it learns your specialized vocabulary by reviewing some or all of your documents and emails. You can also train individual words and terms, and/or create shortcuts to frequently used phrases, signatures, addresses, etc. It's very cool.

I do use a handful of speech recognition phone apps and browser plug ins, along with the native speech to text on my Apple devices...I depend on them to do a lot for me, in fact.
(Edited)
Photo of tony.lima

tony.lima

  • 35 Posts
  • 4 Reply Likes
Thanks.  Maybe I'll give Dragon another try.  I have it but don't use it much.
Photo of Joe Morgan

Joe Morgan

  • 5650 Posts
  • 2929 Reply Likes
It would appear Microsoft is well aware their speech to text engine is terrible.

There's a free Microsoft  Beta "Garage Project" Text to Speech application to anyone that's wants it. http://dictate.azurewebsites.net/

I downloaded and installed myself. It's surprisingly accurate.

 It doesn't transcribe, let you train it words it interprets wrong. But overall, my first impression, it's a winner.

 So this isn't exactly a boon for Camtasia users.But if Microsoft incorporates the same speech engine into their Speech to Text program. Windows and Camtasia users can all reap the benefits. Text to speech was a Camtasia feature that worked so poorly "Not TechSmith's Fault" I always thought that was one of those features that could be removed without offending many people.Maybe Microsoft will resurrect it yet?


(Edited)