Grammar

Nodes and Palette Options > Detailed Palette Option Descriptions > Grammar

	Grammar

Type

Form item

Available from

Any node in which a spoken response from a caller is possible:

AppRoot node (see Setting Global Application Properties: AppRoot Node)

Announce Node

Prompt and Collect Node

Blind Transfer Node

Bridged Transfer Node

Record Node

Form Node

Purpose

An integral part of self-service speech applications is the ability to collect responses from callers and use them to direct the call flow. These caller responses can be in the form of DTMF key presses or spoken commands.

Grammars are used at those points in a speech application where you anticipate a spoken or DTMF response from a caller. Grammars define what words, phrases, or key presses the system can recognize and use as input. Grammar items make it possible to select from among all the grammars defined for an application, to use at specific places in the application.

Note:

In the AppRoot node, a grammar can be used only in conjunction with a Link item. This grammar is active on a global basis. That is, the system responds to any utterance that matches this grammar, no matter where the caller is in the call flow. See Link and Setting Global Application Properties: AppRoot Node.

For more information about grammars, see Working with Grammars.

Behavior

Grammar items cannot be used independently in nodes. Grammar items must always be used as sub-items of:

Bridged Transfer items

Input items

Link items

Record items

In short, you can use grammars only in places where it makes sense to collect and use spoken or DTMF input from callers. You must create the grammar file itself before you can use it in a node.

When a call reaches a node in which there is an active grammar, the system monitors input from the caller. Whenever a caller utterance or key press matches an item in the grammar, the system processes the input according to how the node has been set to respond to that input.

For more detailed information about creating and using grammars, see Working with Grammars.

Properties

Name - Select the grammar you want to use.

Note:

You can select only one grammar per Grammar item. If you want to use multiple grammars at any point, you must use one Grammar item for each grammar you want to use. For example, if you want to offer callers the option to respond either with a key press of 1 or by saying the word "one," you can use two grammars: one to recognize the DTMF input and the other to recognize the spoken response.

Weight - (Default: 1.0) Optional. Enter in this field any floating point number of the format n, n., .n, or n.n, where n represents any length sequence of digits. A weight of 1.0 is considered neutral. That is, the grammar with a weight of 1.0 does not bias the recognition one way or the other. A weight of greater than 1.0 biases the grammar in a positive way. A weight of less than 1.0 biases the grammar in a negative way.

The exact application of weights depends on the ASR engine being used. Therefore, a weight that works well on a particular platform might produce different results on other platforms.

Weights have no effect on DTMF grammars.