Command Windows and enter data — by voice

Topic

New Reply

	BEST SOFTWARE Command Windows and enter data — by voice
	By Patrick Marshall Touch-and-swipe might be the hot new way to work with Windows, but using your voice can be more productive. In an extremely niche category, Dragon NaturallySpeaking is the clear leader. Here’s a look at Version 12.5.

---

The full text of this column is posted at windowssecrets.com/best-software/command-windows-and-enter-data-by-voice (opens in a new window/tab).

Columnists typically cannot reply to comments here, but do incorporate the best tips into future columns.[/td]

[/tr][/tbl]

Reply | Quote

Replies

This was an interesting article, though I must admit that I raised my eyebrows at Patrick’s claim that only one or two corrections per page were needed. I must be doing something wrong! Also I must query his statement that Dragon supports Thunderbird amongst other email clients. Dragon periodically offers its users the opportunity to have it analyse their documents to allow the program to familiarise itself with the user’s writing style and so improve its accuracy. If a Thunderbird user tries to get Dragon to analyse his or her file of “sent” email messages as part of this procedure, it produces a message saying that there are no suitable files to be analysed.

One of the things that Patrick didn’t mention was Dragon’s procedure for making corrections. If there is an error in Dragon’s transcription, the user dictates the command “correct xxxx” (where xxxx represents Dragon’s erroneous choice), whereupon an error box opens with Dragon’s erroneous choice shown at the top and a numbered list of up to nine alternative choices shown below. If the user’s intended word is shown as one of these alternatives, the command “choose (number)” is spoken by the user, and the program is meant to substitute the correct word for the one that it first chose erroneously.

I find three problems with this system.

First, very often the word that I actually intended is first or second in the list of corrections; that is, the program has second-guessed me – and got it wrong.

Second, when one says “choose (number)”, it seems to take Dragon age to act on the command, which one would think it would do almost instantaneously given that it “knows” what’s coming and has only to choose between a very limited number of options.

Third – and quite indefensible in my view – it’s not good at handling corrections for homophones. Supposing one dictates the word which sounds as . This sound could represent , or . The program can’t be expected to know which of these three was intended by the user, so there’s clearly a 2 to 1 chance that it will make the “wrong” selection. The programmers should surely have realised this, and you’d think they would have made sure that the other two possibilities were offered as alternative choices in the correction box. But did they? Of course not!

Reply | Quote

Good article. I used to use Dragon years ago but found it was then mostly only useful for long form, like letters. I stopped using it when I had a system upgrade. (all the sw upgrade costs) However, I did continue using the headset that came with it as it was mic only, no earphones. Nice to hear how its progressed and indeed I know a few people with wrist issues that use it.

Also nice to hear about the Snowball. I just got one for doing podcasts. So much better designed than most USB mics. The non-ICE model I got has the dual capsule so I can change the pick-up pattern for an interview. But yes, the sound is great. Found mine in a Staples clearance bin.

Reply | Quote

I too was a bit surprised about the claim that Dragon speak makes a few errors. Dragon speak makes a lot of errors! I was working on my graduate papers using Dragon speak and believing the advertisements, I did not check my papers and lost a lot of points. I will agree that once you begin to use Dragon speak, it is very hard to put it down. I have been using it since before version 10 and have not missed an upgrade. I do feel that I am better with it that without it but I am not going to beat the marketing drum and say that it is 100% accurate.

I actually upgraded my rig to 16+ megabytes of RAM to accommodate Dragon speak because it is a resource halt (hog) and it did not make that much difference although using Dragon speak with less than 4 GB of RAM really impacts performance.

If you are dictating two or three short paragraphs, Dragon speak will make less errors than if you were dictating two or three pages consistently. In addition, the word processor or editor makes a lot of difference. As we know, Microsoft products are not the most lean and sometimes using libre 4.0 or open office is actually better.

To be certain, Dragon speak will tax the resources on your machine to the Max. My big caution is that you have to have enough resources such as RAM and processing power and in a sufficient video card and running it on something like I three (i3) processor is asking for major performance hits.

I except that speech recognition is still not quite there and not quite there by a long shot. There were a few times where I was about to ask for refund because it was just so hard to get the thing configured but finding the right person in tech support, as usual, did the trick for me. I am running version 12.5 and because I have so many resources on my rig, I can take full advantage of most of these features.

However, the average user should not be fooled into thinking that this is a foolproof application – not by a long shot!

Reply | Quote

I understand Dragon NaturallySpeaking (Premium here) is a first person oriented product, i.e., you train it to recognize how you specifically enunciate. My question is can it be used to recognize someone else’s speech without that person creating their own training files? Usually you read a predetermined script to DNS and it compares your speech to what it expects to hear. Is there any way to get DNS to use a custom script? Then you could take an audio recording, make your own transcript for a DNS script and submit both to DNS for training. My goal here is to transcribe a broadcast personality.

Reply | Quote