 |
|
|
General Application Processes
Speech Recognition Foundation:
All voice applications are designed to guide the user through a series of questions and answers. Pre-recorded prompts are designed for ease of use and for directing the user to respond in the way that the application is prepared to respond accurately. When a user responds to a prompt, either a single word is matched, or multiple words are matched against a state-specific grammar (a "state" can be considered as a single step in the user workflow). The primary grammar represents those words that are possible choices for a given state. The secondary grammar represents words the user may say that are not meaningful to the prompted query, and will first be matched against the grammar then discarded by the application. Any given grammar may have between 20 words (for simple yes/no questions) and 100,000 words (for product catalogs and name databases).Simply Speaking: AnyTransactions re-uses, builds and maintains grammars for each state a user experiences throughout the voice application flow. All words spoken by the user are matched against the application specific grammar supplied to the Nuance Speech Recognition engine. The Nuance Speech Recognition engine uses a North American dictionary (for US English language applications) that is made up of years of real-life experience, covering all major dialects found in the US.
|
Text to Speech Foundation:
Most voice applications will have elements of text to speech used to generate spoken words from purely digital sources. Pre-recorded words and phrases are typically created for common and frequently used information. For variable information that is originates from a voluminous data source (such as all streets and addresses in the US), human sounding, yet computer generated text to speech is used to provide responses. Most voice applications have very little text to speech, and some have none.Simply Speaking: AnyTransactions helps its customers select voice talent appropriate for the application for all pre-recorded responses. The Nuance Text to Speech engine is used for instances where the this function is required.
|
|
|
|
|
|
|
Primary Application States
User Identification:
Names (first name and last name), phone numbers, email addresses, employee numbers, and other codified methods of identification can be used to identify a first time visitor or a repeat visitor to a speech recognition application. Simply Speaking: AnyTransactions uses its pre-developed SpeechObjects capture names, phone numbers (or any numeric string), and email addresses
|
User Authentication:
For applications that require secure access, passwords and PIN numbers can be used. For added security, voice print authentication is available – where a biometric measurement of the user’s vocal utterances is measured and matched against a user profile.Simply Speaking: AnyTransactions uses its pre-developed SpeechObjects for numeric strings (PIN numbers) and alpha/numeric strings. Nuance’s Speech Verifier is used for voice print authentication, as customer’s application require
|
Address Capture:
For catalog, e-business and e-commerce transactions, address capture is critical for both credit card capture and for shipping address capture.Simply Speaking: AnyTransactions uses its pre-developed SpeechObjects to capture address information (city, state, street, zip code) for homes and business located in the United States. This is done by using the zip code as the key to selecting a manageable grammar (meaning not too large) for matching against the user utterance. Alternately and in concert with the Address Capture SpeechObject, ANI (Address Notification Information) can be used for calls that largely originate from home phones, and a 60-75% success rate can be achieved in providing calling number address information – enabling the voice application to compare what the user says to a very specific, transaction-time, dynamic grammar.
|
Transactional Processes:
Every voice application has a highly custom workflow, dependent on both what is being done (ordering product, checking status of order, entering time card, etc.) and on how the existing IT systems are designed to accommodate such business processes. Thus, a voice application is necessarily customized to the business specific workflow and back-end automation processes.Simply Speaking: AnyTransactions creates a voice interface application by first collaborating with a customer to create a visual design that represents the custom voice application workflow. Then, the VIA Technology generates the voice application, and the voice application is connected with the customers existing IT systems that support the associated business processes. When the voice interface application is complete, the user transactions interface with the customer’s existing system in the same way existing computer-based transactions are handled.
|
Payment Processing:
Many voice applications require a method to capture payment information, typically through credit card information. In addition to needing to capturing name and address, the credit number and expiration date is required to complete a credit card transaction. Simply Speaking: AnyTransactions uses its pre-developed SpeechObjects for capturing names, addresses and credit information for all of its payment oriented voice applications. Minimal cost and work is associated with incorporating these elements into any voice interface application. For applications that have a heavy “repeat” customer base, information from existing online systems is used to “remember” as much about the user as online business practice and policy allow for each individual customer.
|
State Management between Voice Application and Existing IT System:
In order for a voice interface application to work seamlessly with existing automated business processes, the flow of information between the user and the core IT system must be managed as an end to end transaction in the same way as if the caller was using the IT system via a keyboard and computer. Proper state management includes transaction set-up, unfinished transactions, and closure of transactions, and all of the workflow between the user, the voice application and the existing IT systems.Simply Speaking: AnyTransactions VIA Technology greatly simplifies the development, maintenance, monitoring and management of such state-based workflow applications. By reducing the complexity in these areas, the time and cost associated with voice interface applications is reduced, enabling far greater economic benefits to the customer.
|
|
|
|
|
|
|
|
 |
|
|
|
|
|
|
|
|
|
