I found this on google while searching on some information on SAPI identifying phrases. This example shows if there is only one property in the rule. So what if there are 2 or more properties in that rule? How would one go about writing the code for this? I am still confused about SAPI and trying to understand it. Any help is welcome, thanks!
The alternate method is add a property to your list tag/items [you appear to
be familiar with properties], iterate through the property tree to find the
property, and then retrieve the associated item from the property. Note if
this is the only property in your recognized rule, then it is fairly easy to
retrieve the property [no need to navigate the property tree].
For example, you could change your rule to be the following:
<RULE ID="VID_Vcs">
<L PROPNAME="NAME_LIST">
<P VAL="1">Mike </P>
<P VAL="2">Davor </P>
<P VAL="3">Kurt </P>
<P VAL="4">Steve </P>
</L>
</RULE>
Use the following code to retrieve the list item value/phrase text
SPPHRASE* pPhrase = NULL;
hr = cpRecoResult->GetPhrase(&pPhrase);
// Check hr
// Let's assume that you only have one property in your rule, so there is only one property in the property tree.
// ASSERT: NULL == pPhrase->pProperties->pNextSibling && NULL == pPhrase->pProperties->pFirstChild
// ASSERT: NULL != pPhrase->pProperties->pszName && 0 == wcscmp(pPhrase->pProperties->pszName, L"NAME_LIST")
// retrieve the list item index [e.g. 1-4], see VAL XML tags in aforementioned grammar
long lRecognizedListItemIndex = pPhrase->pProperties->vValue.lVal;
// retrieve the phrase text
hr = cpRecoResult->GetText(pPhrase->pProperties->ulFirstElement, pPhrase->pProperties->ulCOuntOfElements, FALSE, &pwszListItem, NULL);
// Check hr
// pwszListItem now contains the recognized list item
//compared to the phrase tag of the dictionary (XML)
if(SUCCEEDED (hResult)) {
if ((pPhrase->pProperties != nullptr) && (pPhrase->pProperties->pFirstChild != nullptr)){
const SPPHRASEPROPERTY* pRule = pPhrase->pProperties->pFirstChild ;
if (pRule->SREngineConfidence->confidence Threshold) {
if ( wcscmp ( L"word one", pRule->pszValue) == 0 ) {
//do stuff here
}
else if ( wcscmp ( L"word two", pRule->pszValue) == 0 ) {
//do stuff here
}
else if ( wcscmp ( L"word three", pRule->pszValue) == 0 ) {
//do stuff here
}
else if ( wcscmp ( L"word four", pRule->pszValue ) == 0) {
//do stuff
}
}
}
}
Alright, so sorry for the wait. I whipped up a simple program that may help you figure out what you're trying to do.
So here's my grammar file:
<GRAMMAR LANGID="409">
<DEFINE>
<ID NAME="LIKE_VAL" VAL="1"/>
<ID NAME="SUBJECT_VAL" VAL="2"/>
<ID NAME="COMBINED_VAL" VAL="3"/>
<ID NAME="EXIT_VAL" VAL="4"/>
</DEFINE>
<RULE NAME="LIKE_VAL" TOPLEVEL="ACTIVE">
<L>
<P>I <O>really</O> like</P>
<P>I <O>really</O> do not like</P>
</L>
</RULE>
<RULE NAME="SUBJECT_VAL" TOPLEVEL="ACTIVE">
<P>ponies.</P>
</RULE>
<RULE NAME="COMBINED_VAL" TOPLEVEL="ACTIVE">
<RULEREF NAME="LIKE_VAL"/>
<RULEREF NAME="SUBJECT_VAL"/>
</RULE>
<RULE NAME="EXIT_VAL" TOPLEVEL="ACTIVE">
<L>
<P>Exit</P>
<P>Quit</P>
<P>Terminate</P>
<P>Deluminate</P>
</L>
</RULE>
</GRAMMAR>
And here's a full program that uses it:
#include "sphelper.h"
#include <Windows.h>
#include <string>
int main(int argc, char* argv[])
{
CComPtr<ISpRecognizer> cpReco;
CComPtr<ISpRecoContext> cpRecoCtx;
CComPtr<ISpRecoGrammar> cpRecoGram;
ULONGLONG ullEvents = SPFEI(SPEI_RECOGNITION)|
SPFEI(SPEI_FALSE_RECOGNITION);
ISpObjectToken* pInputToken;
ISpRecoResult* cpRecoRslt;
HRESULT hr = S_OK;
hr = ::CoInitialize(NULL);
hr = cpReco.CoCreateInstance(CLSID_SpInprocRecognizer);
hr = cpReco->CreateRecoContext(&cpRecoCtx);
hr = cpRecoCtx->CreateGrammar(0, &cpRecoGram);
hr = cpRecoCtx->SetNotifyWin32Event();
hr = cpRecoCtx->SetInterest(ullEvents, ullEvents);
hr = SpGetDefaultTokenFromCategoryId(SPCAT_AUDIOIN, &pInputToken);
hr = cpReco->SetInput(pInputToken, FALSE);
hr = cpRecoGram->LoadCmdFromFile(L"Your_Test_File.cfg",SPLO_STATIC);
hr = cpReco->SetRecoState(SPRST_ACTIVE);
hr = cpRecoCtx->SetContextState(SPCS_ENABLED);
hr = cpRecoGram->SetGrammarState(SPGS_ENABLED);
hr = cpRecoGram->SetRuleState(NULL, NULL, SPRS_ACTIVE);
std::wstring strExit = L"Exit";
std::wstring strExitRuleName = L"EXIT_VAL";
CSpEvent spEvent;
bool isListening = true;
do{
hr = cpRecoCtx->WaitForNotifyEvent(INFINITE);
if(spEvent.GetFrom(cpRecoCtx) == S_OK)
{
switch(spEvent.eEventId){
case SPEI_RECOGNITION:{
WCHAR* strReco = 0;
cpRecoRslt = spEvent.RecoResult();
cpRecoRslt->GetText(SP_GETWHOLEPHRASE, SP_GETWHOLEPHRASE, TRUE, &strReco, NULL);
printf("%ls\n",strReco);
SPPHRASE *phrase = NULL;
cpRecoRslt->GetPhrase(&phrase);
if(phrase){
std::wstring ruleName = phrase->Rule.pszName;
if(strExitRuleName.compare(strExit)==0){
isListening = false;
}
}
break;
}
case SPEI_FALSE_RECOGNITION:{
printf("False Recognition\n");
break;
}
}
}
}while(isListening);
cpRecoGram.Release();
cpRecoCtx.Release();
cpReco.Release();
::CoUninitialize();
printf("Press any key to continue...");
getchar();
return 0;
}
You'll have to change the path of where the load grammar call is loading from. From what I understand what you're attempting to do is create grammar in a context free grammar file AND try to do this programmatically as well. Typically you start with a grammar file and modify when you need to.
If, however, you REALLY REALLY need to add new grammars programmatically, such as when someone's typing in new grammar to be recognized, THEN you'd chance SPLO_STATIC to SPLO_DYNAMIC and start implementing the code you see in the later half of the MSDN post you saw.
I completely left out any error checking. If you need to access other properties of the rule you're looking at, use the pPhrase->GetPhrase(&phrase) area. Other than just the rule's name you can also get it's ID.
