IVR Solutions

This section of our technical library presents information and documentation relating to IVR programming and custom IVR software and products. Interactive voice response phone systems and are very popular for service and sales organizations, allowing customers and prospects to call your organization anywhere in the country. Our PACER and Wizard interactive voice response systems add another dimension to our call center phone system solutions.

What Is IVR?. An Interactive Voice Response (IVR) processes inbound phone calls, plays recorded messages including information extracted from databases and the internet, and potentially routes calls to either inhouse service agents or transfers the caller to an outside extension.

Contact DSC today. to learn more about our IVR services and IVR application development software.

BY BOB VILSOET, AMERITECH

DATABASE INTEGRATION

When integrating databases, there are a host of issues that need to be addressed. Among them are database licensing, data integrity, and performance.

Licensing
Data needs to be accessible from the database. Often, the initial lookup into the database fails because the licensing is not properly set up. This causes extra steps to be taken by the programmer during the coding stage of the application. How the data is accessed needs to be determined. Sometimes there are APIs that need to be used to gain access to data. Sometimes the VRU has to emulate an agent’s screen and enter requests. During the design phase of the implementation there is always the issue of whether to use the IVR’s API or another API. Decisions like this often come down to questions of technical ability, development lifecycles, and out-of-pocket costs.

Integrity
Data integrity needs to be maintained. Typically, with an SQL relational database, two programs, applications, or users cannot modify the same data record at the same time. How often the data gets updated must be understood. For example, if a bank’s host application updates data every night, but callers call in before the update is made to the database, incorrect information could be given to the caller regarding account balances.

Performance
Speed of the updates has its own idiosyncrasies. For example, should a message be played to the caller while the VRU is looking up information on a very slow network? Perhaps a commercial announcement should be made. But should the announcement be interruptible when the data comes back? Should the caller be subjected to long messages if data lookups don’t take as much time? These design questions need to be answered for the programmer to stay on schedule with the project.

VOICE RECOGNITION

Voice recognition logic is among the more interesting software algorithms, but the logic is fuzzy at best. The software must recognize words from people with different accents, pronunciations, dialects, and speech inflections. Although different algorithms and design methodologies exist in the marketplace, the basic premise is worth understanding. Each word that is spoken by the caller is broken down into digital information, with each syllable having its own distinct electronic pattern. Those patterns are matched against the patterns of syllables, and the level of accuracy is scored. The syllables are then put together to do a word match, and the highest score is most likely the word spoken by the caller. If there are multiple matches in the word database that pass the user configurable threshold value, then the caller receives a prompt requesting a clarification. If there is no match with a high enough score, the caller is prompted to input the information again, or another method to get to the information is attempted.

These voice prints can also be used for secure voice communications. It is far more accurate to verify that the same caller with the same voice inflections and accent speaks a password than it is for the software to verify that a caller is making a specific request. An application of this nature is valuable in the banking industry where you only want to give financial information out, or allow updates, or withdrawals. By using a voice print instead of a personal identification number (PIN), the opportunity for fraud is significantly reduced.

When sentences are requested of the caller, certain keywords are chosen. For example, a caller who wants to know their account balance could ask, “Can you tell me how much money there is in my account?” The system will find no matches on many of the words and discard them. However, the software will have matched on keywords such as “money” and “account.” The bank balance will then be returned to the caller. There are significant levels of software involved in any such application. There is an operating system, software that digitizes the voice, an underlying application that does word and syllable searches, database access software — preferably fast — and an application that handles the logic of the call flow. There are different voice and software companies in the industry that have specialized in certain subroutines, or areas of speech recognition. Each of these software programs has procedures that can be called to help simplify the applications development. As time goes by, and as the speech recognition software modules improve and the user interfaces are simplified, call center managers will be able to create, design, and write applications — tasks formerly handled by programmers.

In the future, we will likely see voice recognition software enable a voice response unit to completely emulate a live agent, have a conversation with the caller, and handle a significant number of inquiries, requests, and updates.

Bob Vilsoet is director of systems integration for Ameritech’s Call Center Solutions. Ameritech serves millions of customers in 50 states and 40 countries. Ameritech provides a full range of communications services, including local and long-distance telephone, cellular, paging, security, monitoring, cable TV, electronic commerce, online services, and more. For more information, visit the company’s Web site at www.ameritech.com.