Frequently Asked Technical Questions - SDK

When creating applications for multi-language deployment, ensure that the mode attribute for the grammar tag in all grammar files is set to dtmf. This ensures that the grammar file can be loaded into any engine regardless of the language specified.

Enable the AutoPostBack property of the semantic item that you want to monitor.

No. This control only confirms either area code or a local number.

Work around this problem in one of two ways:

No.

To set the bargein type for a particular Speech Control, do the following:

For example, suppose that you want to set the bargein type of a QA control in your speech application to speech. In the Web.config file, add the following SpeechControlSettingsItem element that has an id attribute with the value QAStyle:

<configuration> 
   <system.web>  
       -
   </system.web>
   <speechControlSettings>
       <speechControlSettingsItem id="QAStyle">
           <qa>
               <prompt>
                   <Params>
                       <speech:Param Name="bargeintype">peech</speech:Param>
                   </Params>
               </prompt>
           </qa>
       </speechControlSettingsItem>
   </speechControlSettings>
</configuration>
				

Next, as shown in the following code, add a settings attribute with the value QAStyle to the speech:QA element of the QA control for which you want to set the bargein type to speech:

<speech:QA settings="QAStyle" id="QA1" runat="server"></speech:QA>

Alternatively, you can add a Params collection containing a Param element directly to the QA, as shown in the following code:

<speech:QA id="QA1" style="Z-INDEX: 102; LEFT: 304px; POSITION: absolute; TOP: 176px" runat="server">
    <Dtmf ID="QA1_Dtmf"></Dtmf>
    <Reco ID="QA1_Reco"></Reco>
    <Prompt InlinePrompt="hello" ID="QA1_Prompt">
        <Params>
            <speech:Param Name="bargeintype">speech</speech:Param>
        </Params>
    </Prompt>
</speech:QA>

Note To set the bargein type globally for all Speech Controls of a particular type, create a SpeechControlSettingsItem element within the SpeechControlSettings element of the Web.config file, and assign it the id value globalStyle (case-sensitive).

Yes, this is expected behavior. For more information, see the topic entitled \"Validating Prompts\" in the SASDK Help documentation.

Verify that one of the following two registry keys exists in the Web server registry:

Yes, this is expected behavior. If audio is not part of your logging, removing it should improve performance.

No. These values are associated with utterances.

A QA is active only when its semantic item has not been filled. To return to a previous QA, clear out its semantic item value.

Add the following code in your script to loop over all the answers of the QA and clear the value of each semantic item, where myQA is the name of your QA

for (var i=0;i<myQA.answers.length;i++) {
	var theSI = myQA.answers[i].semanticItem;
	theSI.Clear();
}

The previous code loops over the answers of a QA. To loop over the ExtraAnswers or Confirms collections of the QA, enumerate as follows:
     myQA.extraAnswers
     -or-
     myQA.confirms

The value of Reco.Reject is used by the listen object to determine whether to reject the recognition result and send a NoReco exception to RunSpeech. If a recognition is successful, RunSpeech uses the value of an answer's Answer.Reject property to determine whether to reject the answer as a possible match with the recognition result.

The Answer.Reject property is typically used in cases where a Reco object listens for one of a multiple number of candidate answers. In these cases, the Answer.Reject property of each candidate answer sets the threshold value that RunSpeech compares to the recognition confidence value of the associated recognition result. If the confidence value of the recognition result is lower than the rejection threshold value of a particular answer, RunSpeech ignores that answer. In effect, if conditions such as telephone-line quality causes recognition to be poor, RunSpeech evaluates a larger number of answers to determine which one best matches the recognition, whereas if conditions cause recognition to be exceptional, RunSpeech might need to determine only whether one answer matched the recognition.

Rank and confidence are usually (but not always) correlated. Rank indicates the certainty with which the speech recognition (SR) engine recognizes an utterance, while confidence indicates the degree of certainty that the recognition is correct. Rank is one of the factors that the SR engine uses to calculate confidence.

This is not supported. Application Speech Control functionality is encapsulated by design.

For example, the CreditCardDate control appears to contain three internal QA controls. Can I get the id of the QA that is active at a given point in time?

No. Application Speech Control functionality is encapsulated, and as a consequence, the internal QA ids of the controls such as the CreditCardDate control are hidden by design.

There are four SALT interpreters in the Speech Platform. Each uses a different JScript interpreter. The following table lists each SALT interpreter and the JScript interpreter it uses.

The .aspx page on which you want to use the MakeCall control should contain the following declaration, where the value of the CalledDirectoryNumber attribute is the phone number that you want to call. In this example, the control places a call to the phone number 123-4567.

<speech:MakeCall id="MakeCall" style="Z-INDEX: 101; LEFT: 8px; POSITION: absolute; TOP: 16px"
     runat="server" CalledDirectoryNumber="1234567"></speech:MakeCall>

Use code contained in the recognition grammar file. In the grammar file, create a rule in a DTMF grammar that contains a phrase consisting of the pound sign (#).

A comprehensive discussion of how to write effective grammars is beyond the scope of this documentation. However, the following guidelines provide a solid foundation.

The developer version enables debugging and provides support for the SASDK tools environment. It is identified as follows:

<!-- Microsoft Internet Explorer Speech Add-in-->
<object id="SpeechTags" CLASSID="clsid:DCF68E5B-84A1-4047-98A4-0A72276D19CC" VIEWASTEXT>
</object>

The end-user version does not provide these features. It is identified as follows:

<!-- Microsoft Internet Explorer Speech Add-in-->
<object id="SpeechTags" CLASSID="clsid:33cbfc53-a7de-491a-90f3-0e782a7e347a" VIEWASTEXT>
</object>

If you are developing your application using the QA control from the SDK, then the QA will detect the Speech Add-in component and render accordingly. If you are developing your application using SALT directly or if your application is mixed with SALT code (for example, to provide prompt support for multimodal mode), you must configure the IIS MIME types.

No, both versions cannot be installed on the same computer. If one version is already installed, the SASDK installer does not install the other version.

Yes. However, we strongly recommend that for testing, you use a clean computer that is configured with the same environment the user will have.

Use the user agent string that is documented in the SALT 1.0 specification. Following is an example of a user agent string:

Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; T312461; 
Q312461; .NET CLR 1.0.3328; SALT 1.0.1023 3583).

Yes, you can. Detailed information on supporting DTMF and <listen> can be found in the SALT specification under section 2.3.6 "Using Listen and DTMF simultaneously". SALT enables this in two ways: "(i) the disabling of initial timeouts on the other mode on detection of input, and (ii) the automatic cancellation of one mode when the other mode comes to an end."

The SASDK supports only 8 kHz, 8 bit, mono, PCM recording.

Yes. You can add a debugger statement in the code of the prompt function. When the prompt function is executed, the debugger starts.
Note Script debugging must be enabled within Internet Explorer (under Tools | Options).

Pass document.all["TextBox1"].value into the Runtime Value column in the Prompt Function Editor.

Use a <div/> element in the prompt function. For example, suppose that you have recorded the prompt: "you are flying from", and you want to follow this prompt with the name of a city spoken using a TTS voice. Write the prompt function as follows:

{
    Return "you are flying from <div/>" + runtime_city_name_variable; 
}

Alternatively, you can use the <tts> element to specify content that should be spoken using the TTS voice, as in the following example.

{
    Return "you have email from <tts> Joe </tts>";
}

Do the following:

Yes. Manage the sample applications like any other speech application using Speech Application Deployment Services (SADS). See the Microsoft Speech Server Help documentation for information on using SADS.

No.

Yes. For example, your URL could be "http://MyServer/webform1.aspx?&a=1&b=2&c=3". In the code behind, you can use code similar to the following code that outputs "1" to a text box:

String str1 = Request.QueryString['a'];
TextBox1.Text = str1.ToString( );

This mechanism is not implemented. For a workaround, pass a string into a text box on the webform from the client side, and then retrieve the string from the server side.

Words represented in uppercase letters must be included in the TTS lexicon in order to be treated as a single word. To avoid this issue, use lowercase letters to represent the words that you send to TTS. For example, instead of using <tts>EXAMPLE</tts>, use <tts>example</tts>.

This usually is caused by the value of the ConfirmThreshold property. If you use a confirm threshold, and if the confidence score of the returned result is lower than the value of the ConfirmThreshold property, then the SemanticItem state is set to "NeedConfirmation". The result will not be bound to the text box until the SemanticItem is confirmed. To work around this, either:

     Provide a confirm QA.
     - or -
     Change the SemanticItem in the SemanticMap control, and set its binding behavior to "BindOnChanged".

Set the value of the InputMask property of the control so that it accepts text. For example, if the InputMask property is set to AAADDD, the control recognizes a series of 3 letters followed by 3 digits, such as "a. b. c. one two three".

Insert the following SALT tag, as shown in this example:

<speech:listen id="...." ............. >
    <speech:param name="server">http://Server_Name/SpeechServerWeb/Lobby.asmx</Speech: param>
</speech:listen>

Create a reference to the .wav file using the <content> element, as shown in the following example:

<html xmlns:salt="http://www.saltforum.org/2002/SALT"> 
    <body onload= "prmt.Start()">
        <input id = "ButtonCityListen" type = "Button" onClick = "prmt.Start()" />
        
        <!--SALT-->
        <salt:prompt id="prmt">
        
            <salt:content href="./1.wav"></salt:content>
        
        </salt:prompt>
    
    </body>
</html>

Set the BargeInType value in the Web.config file. Insert the following code inside the <configuration> element of the Web.config file.

<SpeechControlSettings>
    <speechControlSettingsItem id="globalStyle">
        <QA>
            <Prompt>
                <Params>
                    <Param Name="bargeintype">grammar</Param>
                </Params>
            </Prompt>
        </QA>
    </speechControlSettingsItem>
</SpeechControlSettings>

Set the DisableColumnNavigation property to true.

Yes. Using SALT markup, adjust the volume using the volume attribute of the prosody element, as shown in the following example:

<html xmlns:salt="http://www.saltforum.org/2002/SALT">
<object id="SpeechTags" CLASSID="clsid:DCF68E5B-84A1-4047-98A4-0A72276D19CC" VIEWASTEXT></object>
<?import namespace="salt" implementation="#SpeechTags" />
  <body onload="promptHello.Start()">
    <salt:prompt id="promptHello">
        <speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xml:lang="en-US">

            Hello. <emphasis> Hello. </emphasis> <prosody volume = "20"> quiet </prosody>

        </speak>  
    </salt:prompt>
  </body>
</html>

With QA inline prompt:

<speak version="1.0" xmlns:ssml=http://www.w3.org/2001/10/synthesis xml:lang="en-US">
    <prosody volume = "10"> quiet text.</prosody>
    <prosody volume = "80"> loud text.</prosody>
</speak>

The promptSelectFunction functions used in the SASDK samples are contained in .js files (not in .pf files). For example, the prompt function for the sample "Navigating Tabular Data" is contained in the PromptHelper.js file associated with that sample.