For the last several months I have been engaged in researching a 5 part training course to teach real estate agents how to use Social Media and Social Networking to enhance their business. It all came to a head this weekend when I started to actually assemble the course from the mountain of data that I had accumulated.
It was my plan to actually dictate the course into my portable digital voice recorder and then use Dragon Naturally Speaking to transcribe it into text. My plan was to take the resulting text and then use portions of it in the outline for the Keynote presentation. Notice that I said plan.
The reality was quite different. The recording went very well. The transcription from the DVR to text failed so miserably that I was sure that I might miss my deadline. (the first presentation of the course is on Tuesday a few days away). The reason that Dragon failed me is that I had failed to actually train the software to my voice. Without the training it had no chance of turning out any usable text. Of course I spent most of Saturday on a lost cause. Finally by late Saturday evening I decided to do a Google search to find some help. I Googled “Voice to Text Macintosh”. In the first page of results I found an interesting link to Amazon’s Mechanical Turk. It seems that Amazon has a Beta test going of a system whereby users called Requestors (in this case me) can submit a job request that Amazon’s software then parcels out to humans who do the processing. Exactly the reverse of human to computer interaction!
The process was fairly simple. I uploaded my audio file to http://drop.io (which I covered in a previous blog post here). Then I filled in the requested information. Since my file was 1 hour and 8 minutes long, I figured that it would take a couple of hours to complete and decided to offer $25.00 for completion. Amazon added a 10% fee to that and promptly charged my card $27.50.
At this point I am out $27.50 and have NO idea whether or not I’ll ever see my money or a completed file, but I have spent more than that on frivolous things in the past so I wasn’t too concerned. A few hours later I received notice that the file was nearly finished and it would be completed by my deadline of 10 AM, roughly 12 hours after I submitted it. Woo-Hoo!
So, how was the accuracy? Pretty darn good. I’d say it was upwards of 98% correct. I did use some technical terms in the recording and these were transcribed correctly. Some brand names did not fare so well, but are easily corrected. It turns out that Amazon has a handy User Guide available that I completely missed that would have helped me get things done a little easier, but even without it, I got my file and my project was able to be completed on time.
I got a bit curious as to who the “Turks” were and was hoping that I wasn’t contributing to slave labor or child labor in some part of the world and did some research. I ran across this blog, http://waxy.org/2008/11/the_faces_of_mechanical_turk/ that does a good job of explaining who does this work, and why they might want to do so.
I was also curious as to the origins of the term, mechanical Turk and found this website which gives a short explanation.
I realize that many of you will never need to have an audio file transcribed, but that isn’t all the the Mechanical Turk can do for you. I was particularly intrigued by their offer to create content for your blog or website. I may have to check this out the next time I get writer’s block. But you may want to look into it as a way to get your foot into the blogging world, or as a way to give you a safety net when you just can’t thing of thing to write about
Give it a try and let me know what you think, or maybe you have found a better way!
