Speech Auto Attendant Features

Input Options

SAA supports the following input options:

Spoken name: a caller simply says the name
Keypad DTMF: using the keypad to enter digits
Spoken DTMF: speaking the DTMF digits

Some users may use mixed input throughout the call. Input variations include the use of universal commands (such as "help" or "operator"), retry, no input, re-prompt and barge-in.

Refer to the NuPoint Unified Messaging User Guide for more information about input options.

Caller Line Identification

The Speech Auto Attendant can distinguish the origin of a call based on the Caller Line Identification (CLID) of the caller and then play the appropriate greeting for that call.

Note: Three types of system greetings (prompts) can be recorded for the Speech Auto Attendant: Internal Greeting, External Greeting, and Expert Greeting. Refer to Prompts for instructions to record these greetings.

There are four types of callers:

Internal trusted caller: This type of caller places a call from inside the company’s telephony network and the caller ID is recognized as an Auto-Attendant user. The system then plays the trusted greeting (the "bingbing" earcon prompt).

External trusted caller: This type of caller places a call from outside the company’s telephony network and the caller ID is recognized as an Auto-Attendant user. The system then plays the trusted greeting (the "bingbing" earcon prompt).

Internal non-trusted caller: This type of caller places a call from inside the company’s telephony network and the caller ID cannot be recognized as an Auto-Attendant user. The system then plays the internal non-trusted greeting (“Who would you like to call?”).

External non-trusted caller: This type of caller places a call from outside the company’s telephony network and the caller ID cannot be recognized as an Auto-Attendant user. The system plays the external non-trusted greeting (“Welcome. You are speaking to the auto attendant which uses your voice to direct your call. Say the name of the person or department with whom you would like to speak.”)

Note: Barge-In is enabled by default for all three greetings.

All numbers that are configured in the administration web console are trusted. All of the user's desktop directory numbers (extensions) are internal-trusted. All of the alternative numbers under that user (e.g. cell phone, home phone) are also trusted.

All directory numbers (extensions) that do not have a user associated with them (e.g. a phone in a conference room) are all internal non-trusted.

Confidence Thresholds

The system recognizes speech at two confidence levels: low and high.

A low confidence level results when the system does not recognize a requested name, number, or department and instructs the caller to repeat the call request.

A high confidence level results when the system recognizes a requested name, number, or department and thus transfers the call to the appropriate destination.

Confidence thresholds are used to determine whether explicit or implicit confirmation is required. These thresholds are determined by the configurable low/high confidence levels. Refer to Basic Speech Recognition Parameters for confidence level threshold configuration. For troubleshooting help, see Confidence Threshold Recommendations.

Barge-In

This feature allows users to interrupt a system prompt with a speech or keystroke command. The Speech Auto Attendant stops playing the prompt and responds to the command. Barge-in allows experienced system users to skip quickly through the prompts.

Barge-in uses the standard universal commands such as "cancel", "help", and "operator". Experienced users can barge in with voice or DTMF at greetings and questions. Barge-in applies to all three input options.

The system administrator can enable or disable barge-in on a system-wide basis. Refer to Basic Speech Recognition Parameters for configuration.

This single system-wide barge-in configuration is applicable for the following four scenarios:

During initial greetings (any of the three greetings);
During the two levels of re-prompt in cases of no input and recognized result below low confidence;
During the confirmation question of explicit confirmation;
During the prompting of the implicit confirmation.

Refer to the Messaging User Guide for more information about the barge-in feature.

Tutorial

SAA provides a tutorial for new users that allows them to record their name when the Speech Auto Attendant application is configured with a NuPoint Unified Messaging system.

The tutorial is only available to registered internal and external trusted callers. Barge-in will be turned off for first-time callers throughout the tutorial. Trusted callers will be forced to record their name the first time that they call the system. First-time callers are forced to listen to the tutorial to its completion before making any calls.

SAA detects a first-time user only by the CLID of the user's registered desktop phone.

Presence

SAA provides the ability to determine, and optionally play back, the current presence state of the matched person prior to transferring the caller.

Note: The Presence feature is only supported when SAA is configured to use Microsoft Active Directory as its user data source.

The Release 4.0 Presence feature supports the Microsoft Live Communication server (LCS 2005 SP2), Office Communications server 2007 , Lotus Sametime 8.0, and Mitel Unified Communications Server.

The caller is prompted with the current presence status of the person called. The presence information is for reference only, and calls are transferred regardless of the presence status of the called party. Standard universal barge-in commands such as "cancel" and "help" during a call transfer are supported.

Presence status information is optional. It is enabled by default for both internal and external callers, and is independent of whether the caller is trusted or non-trusted. It can be turned off using the Presence for Internal Callers and Presence for External Callers check boxes in the Web Console. For more information, see the Presence section.

Disambiguation

The Speech Auto Attendant provides the ability for calls to be automatically transferred to the operator for multiple directory matches (for example, two people named "John Doe" in the same directory). Callers will only be notified of duplicate names but not prompted to choose a specific one. Callers will be immediately transferred to the operator and will be asked for more details in order to transfer the call to the right person.

If two or more people have names that sound the same, or are the same, the system transfers the caller to the operator, who can then resolve the ambiguity by asking for the department or location of the person being called.

Custom Pronunciation

The custom pronunciation tool allows administrators to customize name pronunciation. For example, if you want to have a user named Mickey Mouse also recognized as Mickey Rodent, you can add a pronunciation of "rodent" for the word "mouse". When a user calls and says either "mouse" or "rodent", Mickey Mouse will be recognized. Multiple pronunciations can be added for each user. Bilingual pronunciations can also be customized (for example, you can also add a pronunciation of "souris" for "mouse".)

You can add nicknames, such as equating "Bill" or "Will" with the name "William".

You can also add phonetic representations of a user's name. For example, an English SAA system will have a higher match rate for the name "Benoit" when you add a phonetic pronunciation (like "benwah") to its dictionary. See Custom Pronunciations for configuration instructions.

Error Handling

SAA responds to an error by instructing the user to retry the command. After two recognition errors, the system transfers the caller to the operator.

Callers are directed back to the main menu or to the operator as governed by the confidence level of recognition and system errors.

Timeout

SAA times out when it detects no speech or DTMF input, and triggers help prompts to be played to the user. The length of the timeout period is configurable in the Basic Speech Recognition Parameters section.