Monday, March 21, 2011

Speak Into The Microphone

There is a widely held belief that when you start talking to yourself it is the first sign of impending madness. There is an addendum which says that when you start answering yourself then you really are some way down the road to madness. Well just recently I paid money for a software package which requires you to speak to yourself and in some cases to even correct yourself. I am not quite sure where this places me on the road, but I have to tell you that it is quite a fun route to travel. There are a number voice recognition software packages available. I chose Nuance’s “Dragon NaturallySpeaking” and although it is early days Dragon is proving to be a useful and practical piece of software. I am in fact using it to dictate this article.

When I first became involved in the computer industry in the 1980s we learned how computers were going to take over our lives, make breakfast in the morning, wash our clothes, and generally perform the role of helpers. The whole “home robot” thing. Well, computers have taken over our lives, but perhaps not in the way in which those futurists anticipated. I can't help get the feeling that sometimes we are their helpers, we certainly are at their beck and call. We are wrapped around their proverbial little techno fingers, and they have us performing circus tricks such as reading newspapers on a tiny 80 x 50mm screens, sending hundreds of millions of inane text messages to one another every day, and endlessly checking weather update phone apps. So while much of the computer hardware and software around us is calling the tune perhaps Dragon NaturallySpeaking is more in line with the ideas proposed by the futurists of the 1980s. It is one of the helpers of the software world.

Like most pieces of software it has a host of features and options which can be tailored to one's individual needs, but perhaps the biggest learning curve is that requiring me to speak to my computer. I have been getting a number of strange looks from my housekeeper who cannot determine whether I am speaking on the phone, speaking to her, or losing my mind. I think she has settled on the latter. Dragon claim that one can dictate up to three times faster than one can type and experience indicates this to be correct. Certainly in my case, the muscular dystrophy is slowing down my typing speed, particularly for articles which require 600 to 800 words. I also find that the freedom of being able to sit back and articulate my thoughts as they come to me provides a great deal of freedom of expression. In simple terms it means that I can babble away freely in the knowledge that the computer will keep up a lot easier than my typing fingers ever would.

This software allows one to control the functionality of the computer such as opening and closing different software packages and navigating within those packages but in my case I am comfortable doing that via the keyboard and mouse. Where the software really comes into its own is the dictation of larger amounts of text. Dragon is able to handle multiple languages, even dividing English into subcategories, presumably to enable it to decipher the subtle linguistic differences between "the rain in Spain falls mainly on the plain" and "yo mama like whatz happenin in the hood like". Dragon does seem to place some demand on the computer's operating system and there are moments when it appears to want to gather its thoughts, but I would estimate that any desktop or laptop computer built the last two to three years will comfortably have enough firepower to handle Dragon NaturallySpeaking. In terms of additional hardware all Dragon requires, other than the computer itself, is a reasonably good microphone. The room needs to be quiet, or the microphone needs to be close to one's mouth, both of which will contribute to better speech recognition. A radio or hi-fi playing in the near vicinity leads to a breakdown of speech quality as the computer struggles to decide who to listen to, you or the music. Dragon guides you through an initial introductory process asking you to read out aloud predetermined text in order for it to learn your speech patterns and pronunciation. The odd error does creep in but I would say that at the moment I am achieving a 90% success rate while speaking in a natural manner. From what I understand this software package is constantly analysing one's dictation and using this to try to more accurately predict the speech recognition so in principle Dragon should get more accurate every time it is used.

Every now and then I have a “blonde moment” and I forget to turn off the microphone when the telephone rings. I prattle away for a number of minutes with the caller, turn back to my computer and wonder why the short e-mail which I was attending to prior to the call now contains paragraphs of text accurately detailing the contents of my telephone call. Sometimes these dictating software packages take their jobs too seriously!

Of course the one benefit is that one also gets better at talking to oneself. Perhaps with time I might even advance to the point where I'm answering myself! Now please excuse me as I need to get back to my padded cell.....