Alpha-Numeric Input from DTMF and Speech-Recognition

[support]

Hello headpill,

After some tweaking with the inline grammar we have come up with a way for a 90%-100% success rate on matching the letter o vs the number 0. Below is the code that is working for us.

We have successfully recited multiple times through testing the following:
BCD120
ODE120
O0O0OO
WAO20

Code: Select all

<?xml version="1.0"?>
<vxml version="2.1">
  <form id="test">
   <property name="grammarmaxage" value="60s"/>
   <property name="grammarmaxstale" value="25s"/>
   <property name="sensitivity" value="0.5"/>
   <property name="confidencelevel" value="0.2"/>
   <property name="interdigittimeout" value="5s"/>
   <property name="speedvsaccuracy" value="0.9"/> 
   <field name="acct">
    <prompt>Please say your account number.</prompt>

    <grammar type="application/srgs+xml" mode="voice" root="ROOT">
        <rule id="ROOT" scope="public">
          <one-of>
            <item repeat="1-7">
              <ruleref uri="#chars"/>
            </item>
          </one-of>
        </rule>

     <rule id="chars" scope="public">
      <one-of>
       <item>a</item>
       <item>b</item>
       <item>c</item>
       <item>d</item>
       <item>e</item>
       <item>f</item>
       <item>g</item>
       <item>h</item>
       <item>i</item>
       <item>j</item>
       <item>k</item>
       <item>l</item>
       <item>m</item>
       <item>n</item>
       <item>o</item>
       <item>oh</item>
       <item>p</item>
       <item>q</item>
       <item>r</item>
       <item>s</item>
       <item>t</item>
       <item>u</item>
       <item>v</item>
       <item>w</item>
       <item>x</item>
       <item>y</item>
       <item>z</item>
       <item>zero</item>
       <item>1</item>
       <item>2</item>
       <item>3</item>
       <item>4</item>
       <item>5</item>
       <item>6</item>
       <item>7</item>
       <item>8</item>
       <item>9</item>
      </one-of>
     </rule>
    </grammar>
  
    <filled>
     <prompt bargein="false">
      I think you said <value expr="acct"/>
     </prompt>
    </filled>

    <nomatch>
     I'm sorry, I didn't catch that.
     <reprompt/>
    </nomatch>
   </field>
  </form>
 </vxml>

We hope this helps.

Regards,
Plum Support

[headpill]

Ok, I will use this Code and try to see if our Speech Recognition improved or not, also zero is taken as Number 0 or not.
I will let you know.

Thanks Support team!

[headpill]

Hello Support,

I see it is now recognizing zero, and thats significant progress.
However, i see the following properties -

Code: Select all

   <property name="grammarmaxage" value="60s"/>
   <property name="grammarmaxstale" value="25s"/>
   <property name="sensitivity" value="0.5"/>
   <property name="confidencelevel" value="0.2"/>
   <property name="interdigittimeout" value="5s"/>
   <property name="speedvsaccuracy" value="0.9"/> 

Is this the best possible values for properties or you can suggest other properties which you think adding can help improvement.

Like, I read in PlumDev Documentation that for 'speedvsaccuracy', Setting a value for this tag has no effect on Plum DEV.
Is helping something here?

Also, If i am taking voice input from caller for atleast 4 times during entire call which could last upto 15-20 mins, and those inputs are atleast 75-90 secs apart from each other, in that case how <property name="grammarmaxage" value="60s"/> and <property name="grammarmaxstale" value="25s"/> will react?

Is the grammar loaded in memory will cleared and not going to load again when the voice input form will be called?. I am mentioning grammar and properties everywhere i am taking voice input.
Or is there anyway i can force previous grammar to clear and load again, or framework is that robust that i do not have to worry about.

Please let me know.

[support]

Hello,

speedvsaccuracy will not affect the values on Plum dev.

It would also be best to remove the property names for grammarmaxage, and grammarmaxstale. By removing these variables, this removes the possibility of caching grammars.

There is also no need to worry about the memory, as that is managed on our end.

Regards,
Plum Support

[headpill]

Hello Support,

The Grammar that you provided works like a charm for an ID upto 4 AlphaNum, anything longer than cuts me of right after 4th char or sometime 5 char. Also, the Zero is fixed but now most of the time i pronounce O it treats it as OH, like for id "OUCH" it speaks back "OHUCH", so i removed the <item>oh</item> and it fixed that issue but i do not know if removing that is going to cause any adverse effect.

Please help.

[support]

Hello headpill,

Removing the <item>oh</item> will be fine. As long as 0 is set in grammar as <item>zero</item> and spoken that way you will not run into issues with it.

To solve the issue of being cut off please look into these two properties:
incompletetimeout: http://www.plumvoice.com/docs/dev/voice ... etetimeout

completetimeout: http://www.plumvoice.com/docs/dev/voice ... etetimeout

Hope this helps.

Regards,
Plum Support

[headpill]

I have used "<property name="incompletetimeout" value="5s"/>" and so far so good, however i do not see much difference in the Explanation of incompletetimeout and completetimeout, both places the description is same.

From this example, the user has 5 seconds to say either one or two. If the user does not say anything within these 5 seconds, then a completetimeout will occur and return a <noinput> to the user. When a user begins to say either one or two, the application waits for 5 seconds due to completetimeout. Once 5 seconds has passed, the application returns a response to the user.

Please let me know if i should use incompletetimeout instead of completetimeout even though incompletetimeout is working for us.

thanks.

[support]

Hello headpill,

We would suggest sticking with incompletetimeout if it is currently giving you successful results.

Regards,
Plum Support