[asterisk-speech-rec] Digit 4, and the digits.gram CRM:00170774

[Stephen Keller]

Ian,

Are you using the built-in LumenVox digits grammar? This grammar uses
our digits-only acoustic model, which does not include "naught" as a
digit since it was built around US English speakers (we have a UK
English model on the way; we hope to have a beta of it available in
about a month).

One thing to try is to make your own digits grammar (using the general
model) that includes "naught" as an option for zero. Right now what is
probably happening is when the Engine hears naught, it matches it with
the closest match in the grammar, which is four. 

If you just used our digits grammar as a model, such a grammar might
look like:

	#ABNF 1.0 UTF-8;
	language en-US;
	mode voice;
	tag-format <lumenvox/1.0>;

	root $Digits;

	$Digit = (ONE:"1"        |
	          TWO:"2"        |
	          THREE:"3"      |
	          FOUR:"4"       |
	          FIVE:"5"       |
	          SIX:"6"        |
      	    SEVEN:"7"      |
	          EIGHT:"8"      |
      	    NINE:"9"       |
	          (ZERO | O | NAUGHT):"0" );

	$Digits = {$=''} ($Digit {$+=$$})<1->;

You may also want to add in other phonetic spellings for "naught" since
I am not entirely sure of all the ways UK speakers might pronounce that
word. See
http://www.lumenvox.com/support/speechenginehelp/programmers_guide/SRGS/
Using_Phonetic_Spellings.htm and
http://www.lumenvox.com/support/speechenginehelp/programmers_guide/SRGS/
Adding_Foreign_Words.htm and
http://www.lumenvox.com/support/speechenginehelp/programmers_guide/phone
mes/EnglishPhonemes.htm for information on working with phonetic
spellings.

An additional solution might be to try applying a weight to zero to make
it more likely to be recognized in the event that the Engine hears
something that sounds similar. See
http://www.lumenvox.com/support/speechenginehelp/programmers_guide/SRGS/
Applying_Grammar_Weights.htm for information on weighting.

I would try the solutions presented in the order I've listed them; our
guess is that just adding "naught" into the grammar will alleviate a lot
of problems you are having.

Please, let us know how it works for you.

--
Stephen Keller
LumenVox Technical Support
P: 877-977-0707, just say "Stephen Keller" or "Support"
F: 858-707-7072
support at LumenVox.com
www.LumenVox.com

LumenVox

Tell us what is important to you! Take our product development survey:
http://www.LumenVox.com/survey/srDevelopment/index.asp




-----Original Message-----
From: asterisk-speech-rec-bounces at lists.digium.com
[mailto:asterisk-speech-rec-bounces at lists.digium.com] On Behalf Of
cyber-cottage.co.uk
Sent: Friday, March 09, 2007 4:39 PM
To: asterisk-speech-rec at lists.digium.com
Subject: [asterisk-speech-rec] Digit 4, and the digits.gram

Hi
 
I am have got one small niggle that is a real pain. My system keeps
reconising the digit 4 as a 0.
 
Basicly what I have a is a simple menu. that passes the call on to have
an account code recognised to check the caller has an account.
 
In the Menu I use the score to check the confidence of the recognition.
if its above 350 I pass the call on, otherwise I pass it back.
Once they get to the account code they are asked to say the code. and
whats happening is that if they say for examle 41234 the dialplan reads
it back to let the caller check it and its often 01230, I cant use the
score as its high normally around 600, and DTMF works fine.
 
Its obviously mixing four up with nought i Guess. Is it possible to
remove nought and make callers say zero ? or does anyone have any other
Ideas ? 
 
Cheers Ian Plain
 
 
<http://brands.xten.net/x-web/LaunchXWeb.htm?SIPSettingsUrl=www.cyber-co
ttage.co.uk/voip/71796.SIPSettings>