In one episode of the original Star Trek television series (Assignment: Earth), the Enterprise and its crew were thrown back in time to Earth in 1968. A young Teri Garr played a secretary that during the course of the episode was astounded by numerous examples of future technology that were being used by an alien from the future. One of these advanced technologies that greatly intrigued me when I first saw the episode in the 1970′s was a typewriter that typed what was spoken to it. In this article, I will refer to the automated typing of speech as “voice type dictation.”
In the late nineties, I experimented with IBM’s ViaVoice software (ViaVoice was sold to ScanSoft now called Nuance in 2003) which promised accurate voice type dictation, but which I found to be woefully inaccurate. I also tried early versions of Nuance’s Dragon Dictation, but found the program to be error prone and more trouble to use than just typing conventionally. However, as time passed voice type dictation software has gotten much better and computing power has greatly increased, in turn increasing accuracy and allowing for more complex processing.
A few years ago, I purchased a PC with the much maligned Microsoft Vista operating system. It came with an installed voice recognition package that was surprisingly good. This prompted me to try the Dragon Dictation product current at that time, Nuance’s Dragon Naturally Speaking 10 (DNS 10), becasue it was reputed to be even better. After some fairly extensive training (reading selected passages to the computer), I found that DNS was quite accurate, once I replaced the cheap mini-plug microphone headset included with the program with a higher quality USB wired headset. I have since upgraded to DNS 11 and a Plantronics Bluetooth USB earpiece microphone.
DNS 11 is more user friendly and requires less training than DNS 10. While I have not been able to achieve the claimed 99% accuracy, I have found that DNS 11 is quite good with accuracy in the 90% range. I can dictate routine letters, emails, etc. about 20% faster than I can type them. However, diligent proofreading is mandatory. I have also found DNS 11 to be less efficient than typing for projects where there are extensive case citations, lots of proper names, acronyms, etc., because these non-standard terms have to be taught to DNS 11 (entering the word manually and then speaking the pronunciation) software or individually spelled out. My primary use for DNS 11 is to dictate time entries. These tend to be repetitive and allow for the use of shorthand macros that save considerable time.
If you have a Windows based PC running Windows 7, you might want to save the $100 or more that Dragon 11 will set you back and use the speech recognition package that comes installed as part of the system. While the Windows 7 system does not have all of the bells and whistles that come with DNS 11, I found it to be quite accurate when a high quality headset is used.
There is also a Dragon application for the iPhone and the iPad that is free and available from the Apple App store. I have found it to be more accurate than the DNS 11 product installed on my PC hard drive, except where it comes to words such a proper names, etc., that the program needs to be taught– there is no function in the Dragon iPad/iPhone app that allows for teaching (other than the option to uplaod your contacts). The application is most useful for dictating short passages for memos and emails.
[On a side note, my Plantronics USB Bluetooth microphone earpiece works well on the iPad and iPad 2. I connect the USB receiver via the Apple USB camera connector (see discussion in last weeks Rubber keyboard article). This earpiece microphone set up is not only good for dictation, but also for making Skype calls on iPad and iPad 2, and for using FaceTime on iPad 2.]
One concern I have with the Dragon Dictation App is that transmits the dictation via the Web to Nuance where it is processed and sent back to the iPad as text. There is also the previously noted option to upload your contact information to increase name accuracy. While this explains the application’s high accuracy level, it may present confidentiality concerns to the extent that a lawyer dictates confidential client or work product information. The License agreement for Dragon Dictation provides luke warm comfort in this regard, indicating that the speech data will be kept confidential, but may be used and shared with third parties for product enhancement purposes only:
“3.3. NAMES AND SPEECH DATA.
(a) NAMES. As part of the Service, Nuance collects and uses the contact names that appear in your address book in order to tune, enhance and improve the speech recognition and other components of the Service, and other Nuance services and products. You may elect at any time to prohibit Nuance from collecting the contact names, via the settings in the Software, at which point, Nuance will delete all contact names that it may have otherwise collected from your address book. Unless you elect to opt out, you acknowledge, consent and agree that Nuance may collect the contact names that appear in your address book in order to tune, enhance and improve the speech recognition and other components of the Service, and other Nuance services and products. Nuance will not contact you or any of the contact names that appear in your address book for any reason, nor will Nuance share your contact names with any third party.
(b) SPEECH DATA. As part of the Service, Nuance also collects and uses Speech Data, as defined below, to tune, enhance and improve the speech recognition and other components of the Service, and other Nuance services and products. In accepting the terms and conditions of this Agreement, you acknowledge, consent and agree that Nuance may collect the Speech Data as part of the Service and that such information shall only be used by Nuance or third parties acting under the direction of Nuance, pursuant to confidentiality agreements, to tune, enhance and improve the speech recognition and other components of the Service, and other Nuance services and products. Nuance will not use the information elements in any Speech Data for any purpose except as set forth above. “Speech Data” means the audio files, associated transcriptions and log files provided by you hereunder or generated in connection with the Service.(c) Any and all information that you provide will remain confidential and may be disclosed by Nuance, if so required, to meet legal or regulatory requirements, such as under a court order or to a government institution if required or authorized by law, or in the event of a sale, merger or acquisition to another entity by Nuance.”
Dragon Dictation License Agreement (emphasis added).
The bottom-line: keep confidentiality concerns in mind whenever you transmit data to third-parties for processing, storage, transmission, etc. This is not a concern for the dictation programs that process the dictation on your computer. However, mobile systems that rely upon processing via the Web must be vetted to ensure that the data is kept confidential and adequately secure to meet the lawyer’s ethical obligations. This is particularly the case for mobile devices such as the iPad, Xoom, Tab and any other mobile device that relies heavily on “the Cloud” for computing power.
In this regard, a good voice type dictation program for the Blackberry is Vlingo, which is available from the Blackberry App Store. I use it for dictating short non-confidential emails and web searches. I also use it for phone number look up for my contacts. In my experience it is accurate, so long as the dictation is fairly simple. Like the Dragon App discussed above, your dictation is processed at a remote facility, so the same confidentiality issues apply.
In summary, accurate voice type dictation is finally a reality and can be a useful tool for lawyers, especially on mobile devices where typing may be difficualt or ackward. However, where the program relies upon third-party processing, keep confidentiality issues in mind when deciding if the use of the application is appropriate.