Grammar slot type

With the grammar slot type, you can author your own grammar in the XML format per the SRGS specification to collect information in a conversation. Amazon Lex V2 recognizes utterances matched by the rules specified in the grammar. You can also provide semantic interpretation rules using ECMAScript tags within the grammar files. Amazon Lex then returns properties set in the tags as resolved values when a match occurs.

You can only create grammar slot types in the English (Australia), English (UK), and English (US) locales.

There are two parts to a grammar slot type. The first is the grammar itself written using the SRGS specification format. The grammar interprets the utterance from the user. If the utterance is accepted by the grammar it is matched, otherwise it is rejected. If an utterance is matched it is passed on to the script if there is one.

The second is part of a grammar slot type is an optional script written in ECMAScript that transforms the input to the resolved values returned by the slot type. For example, you can use a script to convert spoken numbers to digits. ECMAScript statements are enclosed in the <tag> element.

The following example is in the XML format per the SRGS specification that shows a valid grammar accepted by Amazon Lex V2. It defines a grammar slot type that accepts card numbers and determines if they are for regular or premium accounts. For more information about the acceptable syntax, see the Grammar definition and the Script format topics.

<grammar version="1.0" xmlns="http://www.w3.org/2001/06/grammar"
         xml:lang="en-US" tag-format="semantics/1.0" root="card_number">

   <rule id="card_number" scope="public">
        <item repeat="0-1">
            card number
        </item>
        <item>
            seven
            <tag>out.value = "7";</tag>
         </item>
         <item>
            <one-of>
               <item>
                  two four one
                  <tag> out.value = out.value + "241"; out.card_type = "premium"; </tag>
               </item>
               <item>
                  zero zero one
                  <tag> out.value = out.value + "001"; out.card_type = "regular";</tag>
               </item>
            </one-of>
         </item>
   </rule>
</grammar>

The above grammar only accepts two types of card numbers: 7241 or 7001. Both of these may be optionally prefixed with “card number”. It also contains ECMAScript tags that can be used for semantic interpretation. With semantic interpretation, the utterance “card number seven two four one” would return following object:

{
    "value": "7241",
    "card_type": "premium"
}

This object is returned as a JSON-serialized string in the resolvedValues object returned by the RecognizeText, RecognizeUtterance, and StartConversation operations.

Adding a grammar slot type