Tom,
You are quite correct in your assumption about the Version 12 speech engine being a bit more accurate than Version 11. The lag time (the time it takes for spoken words to appear as written words) has been improved and BestMatch IV has been introduced.
BestMatch III was utilized in NaturallySpeaking 10 and consists of trigrams. Essentially, if you dictate a minimum of 5 words, NaturallySpeaking can check 2 words to the left and 2 words to the right of the 3rd word of the phrase. NaturallySpeaking 11 introduced BestMatch IV which is based on quad grams. BestMatch IV comes into play when you dictate a minimum of 7 words. BestMatch V is exclusive to DNS 12 and comes into play when you dictate a minimum of 9 words.
Your advice about the microphone is also spot on. The enduser should have a USB microphone whether the souncard is fused onto the mic or the mic is connected to an external USB soundcard as the integrated soundcards on even high end computers produce less than optimal results. One doesn?t have to buy the most expensive mic on the market but should purchase a mic tested and approved by a reseller that specializes in them.
The computer is also very important. Nuance lowballs the specs one should have to run DNS effectively, we like to see our clients have 8 GB of RAM and really like 16 GB. When a client comes to us wanting to buy a full copy of Pro or Medical with less than 4 GB of RAM we actually turn them away and advise them to upgrade their RAM or computer before rushing into a purchase from us. The best place to go for detailed PC talk in relation to DNS is The KnowBrainer Speech Recognition Forums computer sub forum
http://www.knowbrainer.com/forums/forum/index.cfm where you can find up to date active conversation. There?s actually quite a few docs that participate on there as well with personal posts numbering in the hundreds.
