Feature

Digg This!

While most companies today are under tremendous pressure to anticipate and respond to customer demands in order to stay ahead of the competition, many are meeting this challenge over the Web with enterprise applications deployed on application servers. Increasingly, businesses are being challenged to deliver consistent and immediate services across multiple communication channels to customers who are on the move and whose needs are constantly evolving. To meet this challenge, companies must offer multiple access points to customers - and employees, suppliers, and business partners - in order to provide superior service and win customer loyalty. Web-accessible enterprise applications are a key part of this strategy. Accordingly, there is a huge need for companies to integrate voice-enabled services into many of their Web-accessible enterprise applications. Building separate systems for the various delivery channels is cost-prohibitive and doesn't adequately address the business needs of most companies. The integration of voice into Web services through the existing application server allows the creation of integrated Web and voice self-service, a new capability offering major benefits for cost containment, revenue enhancement, and customer satisfaction. Voice-enabled enterprise applications can productively and cost-effectively:

• Automate complex business processes and repetitive transactions and services.
• Extend customer reach by giving everyone with a phone, wireless, or wire line 24x7 access to enterprise applications.
• Improve contact center resource allocation and utilization by migrating agent tasks to transactional applications using a conversational interface to existing applications.
• Deploy multilanguage support of business processes by enabling callers to speak in their native language.
• Broadcast alerts, notifications, and reminders of events, meetings, prod- uct shipments, account status, etc.
• Converge phone-based voice transactions with Web-based transactions using existing technology, infrastructure, and business logic.

Voice self-service enables companies to meet the requirements of today's "never-satisfied" customers by providing anytime, anywhere access to enterprise information such as order management systems, bill and payment services, inventory and scheduling, customer service, and account management. Compared to menu-driven touchtone systems, customers and employees can expediently and intuitively manage transactions on their own over the telephone by resonding to system prompts. It promotes personalized customer service - including targeted up-selling, cross-selling, and discounts and promotional offers - based on customer profiles created through spoken input.

Fueling the growth of voice self-service in the enterprise is the realization that even with the Internet and Web-based services, the telephone is still the primary way for customers and employees to contact a company. While the number of Internet subscribers continues to rise, projected increases in the number of mobile phone users will only add to the call volumes that corporate contact centers can expect in the future. For example, Gartner Group estimates a 30% increase in calls to contact centers by 2003, when 45% of contact centers will utilize automated voice recognition-based user interfaces.

With the role of voice self-service in the enterprise destined to grow, voice infrastructure software vendors are providing Web/Java developers with powerful VoiceXML and J2EE tools for designing and deploying next generation voice-enabled enterprise applications. These services, deployed on top of existing application server infrastructures, extend a company's reach from the existing Web-based user to the traditional telephone user. They also allow companies to leverage their investment in traditional Java developers to provide these services without hiring additional, specially trained, voice-centric or call-center developers.

Self-Service Infrastructure in the Enterprise
Many corporations invested heavily in Web technology because of its promise to improve operating efficiencies, contain costs, and increase revenues. Today, companies are sitting on top of an enterprise application infrastructure including databases, application servers, and application software, all designed to drive one presentation layer - the Web.

What if this infrastructure could be extended to drive services across the communication channel that's responsible for the vast majority of customer contacts - the telephone? Voice infrastructure software closes the gap between the phone and the Web, as shown in Figure 1, by providing a platform to develop, test, deploy, and manage voice applications that implement a voice user interface (VUI) to a company's existing Web-based applications and enterprise systems. The application server thus becomes both a broker of Web-based transactions and an interface for the high volume telephone customer.

Voice Infrastructure Software and J2EE
State-of-the-art voice infrastructure software architected on J2EE (see Figure 2) and deployed on existing application servers like IBM WebSphere speeds and simplifies all phases of voice application development, integration, and deployment. Businesses can extend their existing technologies and enterprise services with voice services that leverage:

• A J2EE-based services framework for back-end system integration
• Business logic and data across applications, eliminat- ing the need to re-create or store processes for mobile or voice services
• Existing application server investments
• Existing enterprise/Web/Java developer skill sets

A complete voice infrastructure platform allows easy integration of voice with J2EE-based e-business applications and enterprise software. It provides a developer workbench comprising essential voice application development tools:

• VoiceXML development tool: For point-and-click authoring of VoiceXML applications
• Reusable, extensible Java dialog components: For creating voice user interfaces
• Application server-based repository: For storing and retrieving voice assets such as prerecorded prompts, pretested grammars, and VoiceXML and JSP scripts
• J2EE interfaces: For interfacing and communicating with the existing application infrastructure
• VoiceXML debugger: For accurate line-by-line debug- ging of VoiceXML scripts
• Virtual telephony client with headset PC microphone and speakers: For rapid testing and debugging of VoiceXML scripts in a desktop PC environment
• Event tracker and rules engine: For voice user inter-face personalization and dialog stream analysis

Once development and testing of a VoiceXML application is completed, the voice asset repository enables rapid deployment to WebSphere and supports integration with enterprise content management systems such as Interwoven.

For deployment, a voice gateway is needed to enable telephone access to enterprise applications. This gateway consists of a VoiceXML interpreter that allows the integration of speech recognition, text-to-speech, and telephony technologies in an application. The VoiceXML Interpreter interprets and executes the VoiceXML dialog flow of the application by playing, appending, and concatenating the appropriate recorded prompts and relevant text-to-speech content. In addition, the VoiceXML Interpreter processes user requests for information and services by capturing the user utterances and requesting and processing speech recognition results. For all of this to integrate seamlessly, the voice gateway must be able to communicate efficiently with the back-end services, specifically those already deployed on existing application servers.

In the deployment environment shown in Figure 3, businesses can deploy and manage innovative, dynamic, personalized, conversational voice applications that provide self-service over any telephone. Personalization of voice applications is supported by an event tracker loaded onto the application server at deployment, which gathers user statistics and utilizes them within a VUI rules engine to modify voice prompts and options delivered to a user. Remote deployments to distributed gateways and enterprise platforms can be coordinated from a central location, allowing multiple access points to the same application base over the Internet. Deployed in this manner, voice application communication can utilize existing Internet infrastructures without the overhead of Voice over IP or other more expensive solutions.

Java Dialog Components
Predefined Java dialog components simplify the development of natural conversational voice user interfaces. Each component represents a pretested unit of conversation for a commonly used dialog function, and is accompanied by a grammar and prompt library and a parameter set for extending and customizing the dialog function. Dialog components can be extended and reused for new purposes. A parameter set facility enables unlimited extensibility of the dialog functions while maintaining consistency of the voice user interface.

Grammars and prompt assets associated with each dialog component are stored in the voice asset repository. They include:

• Prerecorded audio files: .WAV files that contain recorded prompts
• Prompt components: Scripted, reusable, customizable prompt components that provide quick and easy assembly of audio files
• Prompt pools: Collections of prompts that are randomized to allow the voice user interface to sound more natural
• Parameter sets: Predefined sets of parameters that tailor the component for a particular dialog function
• Grammars: Speech recognition grammars used as the rule base for recognizing words spoken by the user

Personality and application context-specific information can be associated with any prompt via metadata to enable the design of high-quality voice user interfaces. An enterprise application running in an instance of WebSphere manages all of these assets.

Integration of Voice Applications
A state-of-the-art voice infrastructure allows easy integration of voice applications with J2EE-based e-business applications and enterprise software (see Figure 4). Voice applications can request services and send messages to other applications and legacy systems using standard Java Messaging APIs including:

• JMS: For performing point-to-point or publish/subscribe messaging with external resources
• JAXM: Communicating with external resources using Simple Object Access Protocol (SOAP) 1.1
• JavaMail: Accessing standard e-mail services
• JAAS: Authenticating and enforcing user-access controls

A well-designed J2EE voice infrastructure provides integration services that save considerable development time by enabling straightforward integration with back-end systems. This can be done through a set of dialog components called directly from the VoiceXML application that map to the J2EE services:

• Call component: Takes a message and sends it to the specified destination using either JMS or JAXM. The component can be invoked in both synchronous and asynchronous mode. In asynchronous, a VoiceXML script can be called each time a response is checked for, allowing the application to prompt the caller regarding the progress of the request.
• Send component: Takes a message or object and sends it to the specified JMS queue or topic. It can perform this action in synchronous or asynchronous mode. In asynchronous mode a script can be invoked each time a response is checked for. The response in this case would be a JMS message acknowledg- ment to the voice application.

• Login component: Accepts a numeric user ID and optional numeric PIN and passes the information via JAAS to a back-end validation mechanism.
• BrowseList component: Allows the user to navigate through a list of items and perform an action. Supports back-end integration with database items and e-mail via JavaMail.

The J2EE integration services also include voice application service extensions and a messaging controller to enable interaction with the Java Messaging APIs, namely:

• Voice application service extensions: Custom tag libraries that allow server-side applications to incor- porate the same messaging structure that is used in the dialog components. Tag libraries exist for build- ing XML messages and interacting with JMS or JAXM.
• Messaging controller: Provides a mechanism for managing the multiple interactions that can be pres- ent within the messaging services. The Messaging Controller allows JMS messages to be set to wait for an acknowledgment and JAXM messages to wait for the call to complete.

Conclusion
A comprehensive set of VoiceXML and J2EE-based tools gives businesses the solution they need for adding voice services to their enterprise and gaining the benefits of conversational, personalized, voice self-service applications. These benefits include reduced customer service costs, expanded access to enterprise services by customers, consistent information and messaging across all communication channels, enhanced customer satisfaction and loyalty, and new revenue opportunities.

By providing these services through standard Java and J2EE interfaces, enterprises can leverage their existing investment in application server-based Web applications, as well as their investment in the Java and J2EE programming skills they have already acquired to develop those applications. Extending the reach of Web-based applications to the voice market is now an easily attainable, and cost-effective, goal for most WebSphere customers.