MacSpeech Dictate review


Review of the new speech recognition software "MacSpeech Dictate",
version 1.0 offered for Mac users on February 15, 2008

Reviewed by your humble servant Mark Conrad, an old senile
Mac user, who has trouble with such complex things as
finding the "Power On" button on his Mac.

Mac is a MacBook Pro with 4GB ram.
CPU is 2.4 GHz Intel Core 2 Duo
Int hard drive is 150 GB, 2 partitions
Partition #1 roughly 108 GB has Mac OS 10.5.2
Partition #2 roughly 40 GB has Windows OS "Vista Ultimate"


MacSpeech Dictate was installed in the Mac partition (naturally)

Headset microphone and "Translator" (ext' sound card) are both
included in the MacSpeech Dictate software box.

Headset (microphone) is model "VXI USB 7.02" from www.vxicorp.com

USB Translator (external sound card) is from by the same vxicorp.
It is a small encapsulated device that is between the headset
and the USB connector on the Mac.



After I put my loosely fitting false teeth in my mouth,
(they tend to flop and clatter around) - I "trained" the MacSpeech
software using the included training lessons of the software.

Fortunately, my false teeth did not fall out.

After training the speech software, I rounded up some text
that MacSpeech Dictate had never seen, and "spoke it into" my
Mac, in a low volume normal voice, like someone would speak
to his friend, if that friend's ear was a foot away.

Speaking speed was a normal 100 wpm, and all the punctuation
was "spoken in" also, saying "comma" or "period" or "new line"
when required.



ORIGINAL TEXT BELOW, AS READ INTO MACSPEECH DICTATE
**************************************************

The Shiek was his usual resplendent self, grey vest, navy blue
sport jacket, handmade Italian shoes and a Countess Mara tie.
There was the inevitable cigar sticking out of one side of his
mouth, and the thinning black hair and the neatly trimmed grey
moustache was flanked by two smoothly-shaven pink jowls,
campaign ribbons so to speak, from many a boisterous night
on the town in such towns as New Orleans, Chicago, New York,
Nogales, Montreal and Tijuana.





RAW TEXT OUTPUT FROM MACSPEECH DICTATE
*************************************

The sheik was his usual resplendent self, gray vest, navy blue
sport jacket, handmade Italian shoes and a Countess Mara ties.
There was the inevitable cigar sticking out on one side of his
mouth, and the thinning black hair and the neatly trimmed gray
mustache was flanked by two smoothly shaven pink Giles,
campaign ribbons so to speak, from many a boisterous night
on the town in such towns as New Orleans, Chicago, New York,
New Delhi is, Montreal and Tijuana.


Discussion -

Dragon Dictate made three mistakes:

1) "Countess Mara ties" S/B "Countess Mara tie"

1) "smoothly shaven pink Giles" S/B "smoothly shaven pink jowls"

2) "New Delhi is" S/B "Nogales"


There are a total of 79 original words, 3 of which MacSpeech got wrong.
(I did not count borderline words as wrong, such as the variant
spelling "moustache" used in the original text, instead of
the much more common "mustache")


That is approx' 96.2% raw accuracy, darn good for the very first
version 1.0 of MacSpeech Dictate.


I am _really_ surprised that the program got the
"Countess Mara" correct:


Background for the "Countess Mara" term:

[www.CountessMara.com
Luxury Mens Accessores Since 1938 Finest Silks,
Leather Craftsmanship]


Woman whose first name was Mara, born in Rome in 1893.

She originated the idea of the posh tie business in New York,
after betting her husband that she could create better
men's ties than those presently available at the time.



Anyhow, as MacSpeech adds upgrades, this 96.2% raw accuracy
will likely improve considerably.


FINALLY Mac users might have speech recognition which is as
accurate as PC users are used to, thanks to MacSpeech.

Mark-
.