Speech recognition software has come a long way since its first release several years ago and although it is still not perfect, there are many users of speech recognition software that boast accuracy rates of up to 99% but there is some complex programming and data logging required behind the scenes.
The most popular speech recognition programs are Dragon Naturally Speaking and the tool on Microsoft Vista called simply Microsoft Speech Recognition and while both garner many claims about high levels of accuracy , this accuracy comes at a rather high price: extensive training to improve speech recognition and improve accuracy.
Why is Training Necessary or Important for Speech Recognition to Work?
The training process for speech recognition software is required so the computer can learn about the unique inflections and variations in your speech. Even if you speak like a broadcaster in perfect English, the software will not work perfectly out of the box. Everyone needs to train their speech recognition software, some more than others. Training is one of the most critical components of using any speech recognition software. By spending the right amount of time training the speech recognition software, even though it takes quite a long time to do, in the end it will save a great deal of time as you will not need to correct as many mistakes. For those who are only using speech recognition software to take notes or to compose documents that do not require a high level of accuracy, the basic training offered within the core features of the software should be enough.
How Long Does it Take to Train Speech Recognition Programs?
If you have a thick Southern or other accent, you can expect your training to take far longer than it would for someone who sounds much like any of the news anchors we’re all familiar with. By longer, this means you can expect training her speech recognition software to take several months. Even with excellent English speaking skills and broadcaster-like pronunciation, adequate training generally will take around three months of constant use. By constant use, this means that when using the software and training it, you correct mistakes made by the program using the methods the help section dictates. For instance, in both Microsoft speech recognition and dragon naturally speaking, in order for the computer to learn from your mistakes, you must properly correct all errors. Again, this can be a very time consuming process and this training can be very frustrating because so many of us are used to things working straight out of the box. When you begin to use speech recognition software, it is important to look at the process as developing a relationship with both the software in your computer. If you have the patience to perform the training necessary, the accuracy rates will allow you to compose documents far faster and will allow you the freedom to use a headset in your car or other places where you would not normally be able to type.
Generally speaking, if you speak standard American English and enunciate clearly while speaking using a quality headset microphone, over the course of three months of repeated use you can expect accuracy rates for your speech recognition software to be in the 90th percentile. By repeated use, this means nearly every day and always correcting errors using the suggested method by the software. If you have an accent of any kind, the amount of training required to achieve high accuracy rates with your speech recognition software could take between six months and one year. This is a very long time to spend training and the process of training speech recognition software can be incredibly frustrating, to point where you will feel like there’s no end in sight and the computer is not learning from your mistakes. However, for those who what a ability to type because of carpal tunnel syndrome or other problems, this is a relatively short amount of time compared to the alternatives, such as paying a typist or slowly typing things out yourself.
Why Speech Recognition is Worth the Time Spent Training to Improve Accuracy
It should be repeated that training your computer to recognize and interpret correctly the sound of your voice is very much like developing a long relationship. There will be times during the training process that you feel like it is not worth it or that all the training is a waste of time, but get through this frustration because in the end, the benefits offered by speech recognition software are great and you’ll find that the training was very much worth the effort and long training period required to improve accuracy. Do not fool yourself into thinking that you speak great English and do not require extensive training. Instead, expect the training process to improve the accuracy rates of your speech recognition software to take between 3 to 6 months, and for those of thick accents, perhaps as long as one year.
Several people have offered comparisons between Microsoft speech recognition and the popular commercial product called Dragon Naturally Speaking. It appears at the accuracy rates are quite similar in terms of the time required to train the programs is also equivalent but there are minor variations that cause some people to prefer one speech recognition program over the other. Certainly, it is a benefit that the speech recognition software offered by Microsoft comes standard on Cista whereas that offered by Dragon costs significantly more. The choice is yours as to which speech recognition software you choose to use, but whichever program you decide upon, do not fool yourself about this significant period of time required to adequately train your speech recognition software to have high accuracy rates and to work smoothly.