How I Use Voice Recognition

I thought I would use this post to share how I use voice recognition in my practice.  I have attempted to use voice recognition often since the late 1990’s, with little success and much frustration.  I have previously used earlier versions of IBM ViaVoice, and early editions of Lernout & Haupie’s Dragon Naturally Speaking, but every attempt left me more frustrated and a less efficient.  It was not until Scansoft’s Dragon Naturally Speaking version 6 that I became proficient enough, or possibly that the technology became effective enough, to make automated speech useful in my law practice.  I can identify three things that made a significant difference in my effective use of voice software.  First, I obtained a box that would handle, and added 1 Gig of RAM.  Second, I inquired of a high-quality microphone made by Sennheiser.  And third, I purchased approximately 2 hours of online coaching and training with an expert.  (I have sense lost his information, but recall he was from California and had a set of teaching videos).  With these additions, and the purchase of the Law version of DNS 6, I was able to effectively train the software and actually use it in my practice.

I now do all of my dictation using Dragon naturally speaking version 8.1.  I have resisted the urge to upgrade to version 9 of the software, although I do understand it is an improvement.  The two microphones I have found to be most effective are the Sennheiser ME3, which I always carry in my laptop bag for dictation while travelling, and the Plantronics CS 50 wireless microphone, which I use daily in the office as both my Dragon headset and my telephone headset.  I would also note that my colleague and long-time speech recognition user, Chip Westbrook, has implemented a VXI  BlueParrott wireless Bluetooth microphone, with excellent results.  In fact, Chip recently posted and very thoughtful summary of his voice recognition experience to our borrowers technology list served, which I will re-post here as a separate entry as well.

In order to make speech recognition effective, the user must be committed to training of the software by correcting words using the software's correction dialogue wizard.  In the short term, this type of correction takes longer than simply over typing the wrong word, but long term it actually trains the software to understand the way that the user speaks, and significantly increases the accuracy.  Another big time saver is the use of phrases in the software, which can and entire paragraphs of text by simply speaking a couple of words.

All of my correspondence and all longer legal pleadings and briefs are dictated using Dragon.  However, before using the software in this manner, I would caution any user to have a good proofreader follow behind looking for word errors.  The software, while accurate, frequently places the wrong word or tense on the page, although it does not make spelling errors.  As such, is often difficult for the speaker to recognise his own errors when proofreading his own work.

I believe Dragon to be a time saver in my practice and worth the additional effort to learn and train.  In a future segments I will post some additional time saving products that make speech recognition more pleasant and deficient.