Speech Recognition – It Really, Really Works
By Jim Calloway
Interacting with a computer by verbal commands has long been a staple of science fiction from Forbidden Planet’s Robby the Robot to 2001: A Space Odyssey’s HAL 9000.
Speech recognition has improved greatly over the years. It has been a usable tool for some time now. The last few months, it has worked for me much better than ever before. Often a relatively small thing can make a big difference. I decided to cover speech recognition this month because I hope that some of you may benefit and start using speech recognition more effectively yourself after you hear about the small change that made such a difference for me — especially if you are not a speedy typist.
In the 2011 article, I wanted to make readers aware that Dragon NaturallySpeaking 10 was greatly improved over the earlier versions. I also wanted to report on the Dictamus app that allowed an iPhone to function like the hand-held recorders that many lawyers have used, with playback and rewind features. I mentioned that there were several virtual transcription services where raw dictation could be emailed for an individual to transcribe it. I also noted that I had never really liked using headsets, and the Samson Go Mic was working better for me than a headset.
But there have been several developments since then. The Dictamus app (by JOTOMI GmbH) is now available for Android phones in the Google Play Store as well as for iPhone and iPads in the iTunes store. (The Android version is 1.1.6 and has only received 11 reviews to date. So there may be updates ahead.)
Dragon NaturallySpeaking Premium is now in Version 12
. The premium version costs $199.99, but when I checked the website it was on sale for $149.99. I had recently received an email solicitation as a current owner to purchase an additional copy for $99.99. When version 12 is installed and allowed to check for upgrades, it will be upgraded to version 12.5 at no extra cost. It remains my strong recommendation to buy the premium version and not the lesser home edition or the more expensive professional or legal editions.
But my great, new personal upgrade was quite unintentional. Sharon Nelson and I have been recording our Digital Edge: Lawyers and Technology
podcast since 2007. This year our host for the podcast was changed to the Legal Talk Network (LTN)
. The fine folks at LTN wanted better audio quality from my side of the podcast, and after I failed to respond to their hints to buy a better microphone, a package from LTN was delivered to me containing a Blue Snowball microphone
and something called a pop diffuser to place between me and the microphone. The podcast sound quality did improve.
I was quite pleased with my new acquisition — a professional-looking microphone on a tripod. I decided to try using my new microphone with DragonDictate NaturallySpeaking. To me, the results were amazing. First of all, my accuracy seemed to improve. It was already very accurate, but I believe that the larger, higher quality microphone caused some improvement. But the primary improvement was that I was able to dictate more conveniently and so I dictated more frequently.
I had been using my Samson Go Mic. I would park it clipped to the top of my laptop screen and pick it up and hold it close to my mouth to use for dictation, so I often wouldn’t use it for a shorter email (and I often still don’t dictate short emails). When I would use the laptop for a county bar presentation or take it home to do some work, I often would be in a hurry upon my return and not set up the Samson Go Mic, so I would revert back to my old habits.
DragonDictate NaturallySpeaking has this really nice feature — you can leave it on all day. When you don’t want to use the software, you tell it to “go to sleep.” When you want to use the software, you tell it to “wake up,” and you can start dictating. So without having to load software or grab the microphone, I can just turn to my laptop, give the “wake up” command and start dictating. If I get a phone call or just want to think for a moment, I can tell it to “go to sleep” as often as I wish. Your office may be noisier than mine, and this may not work as well for you, but I can now easily dictate with the microphone sitting two feet away from me.
I’m not endorsing this particular microphone or approach. As the saying goes, your mileage may vary. But I’m very grateful to the folks at Legal Talk Network and was actually a bit surprised when a Google search informed me that this microphone only costs about $50.
USING DRAGONDICTATE NATURALLYSPEAKING
First of all, I have been using it for years, and so a lot of the commands are pretty natural for me. But any lawyer should be able to pick them up in an hour or so. For example, when it gets a word wrong, I merely tell it to correct that word and it presents me with a menu of alternatives. Usually the word I intended to use is the first one on the list, and I am able to substitute that word by saying “choose one.” Rarely is the word not on the list, but if it is not, I can spell it and it will add that word to its vocabulary associated with the way I spoke the word.
Training with DragonDictate NaturallySpeaking is pretty simple. You just read a few paragraphs; but it is critical that you allow the software to review your documents and the sent items in your email program on your computer,so that it learns what types of words you use. The very first time I used the word TECHSHOW, Dragon knew to put it in all capital letters. It also got a bunch of scarcely used technology terms correct the first time I used them because it had reviewed my writings on my computer. But I had certainly never used the word “diffuser” before and it got it right the first time when I was dictating this column.
I can dictate a phrase like a book title and then tell it to select the entire phrase and say “cap that.” It will then capitalize the first letters of all the words selected; yet it “magically” knows not to capitalize words like “of”
Two monitors are much better for using Dragon NaturallySpeaking because the toolbar does take up a little real estate at the top of the screen. The Dragon sidebar is also very helpful for beginners, but you will probably want to keep it minimized most of the time as it also takes up quite a bit of space.
Be assured that you will make mistakes and at some point you will forget to tell it to “go to sleep,” and you will find an entire telephone conversation transcribed into the middle of a brief or contract.
If someone else is in the room and they begin talking, you may find that you don’t get any meaningful results at all from dictation.
1) Make backup copies of your individual speech user profiles frequently. We are all familiar with the need for good backup processes. But with speech recognition software, it constantly updates the profile. At some point, if your recognition accuracy decreases greatly, you may just want to restore the profile from a couple of months previously.
2) Make a different profile for every setting. If you have a laptop and take it home to do dictation, you will want to set up a separate profile. Mine is “Jim at home.” The acoustics for different rooms will, in my experience, confuse the software and decrease accuracy. If you find yourself dictating in an unfamiliar or one-time location, then when you close the software, be sure and tell it not to update the user profile.
3) Likewise, if you had a head cold one day or just have a day of poor dictation accuracy, tell it not to update the user profile at the end of the day when you turn it off. There is no reason to update with what may be bad data.
4) Dragon NaturallySpeaking is a resource hog. If leaving it on all the time impacts your computer’s performance, you may need to add more memory or buy a nice new computer.
5) Proofreading a speech recognition-generated document is a bit different than normal proofreading. There are no misspelled words, but there will be a few words inserted that sound like what you said. Sometimes they are harder to pick out.
There is a lot more to this software than I am covering here. I wanted this column to be short and readable.
I do think all of us are going to be using speech recognition a lot more in the future. Many lawyers tell me how they love using Siri to dictate short emails and text messages on their iPhones. I still frequently do voice searches on my iPhone with the Google app, even though my family has requested me not to do that around them in public because “it is just embarrassing.”
The Google Chrome browser will now respond to an “Ok Google” voice command to do Google searches thanks to a new “Voice Search Hotword extension” that’s available to download now from the Chrome web store
. You can also use it to set reminders and do a few other things. It is handy to have a quality microphone easily available for that purpose as well.
I wouldn’t want you to get the wrong idea about my Samson Go Mic. While I do not use it to do dictation anymore, it is still a handy item for my traveling computer bag as a backup microphone for the headset.
But for me the bottom line is that I have a great new “toy” that has improved my productivity – and it’s not even Christmas yet!
If you decide to try the Snowball microphone, feel free to share your results with me by email.
Mr. Calloway is director of the OBA Management Assistance Program. Need a quick answer to a tech problem or help resolving a management dilemma? Contact him at 405-416-7008, 800-522-8065 or firstname.lastname@example.org
. It’s a free member benefit!
Originally published in the Oklahoma Bar Journal -- December 14, 2013 -- Vol. 84, No. 33