Platforms
Promptu's Network Speech Recognition (NSR) architecture integrates the capabilities of handsets, servers and the cellular data network itself to provide seamless, secure, reliable service.
Network Speech Recognition
The NSR architecture divides system functionality among three components: the mobile client, a high-speed server, and the cellular data network itself.
Mobile Client The mobile client comprises the Promptu software that resides on, and runs within, the user's mobile handset. The client serves two functions. First, it interacts directly with the user, both capturing input in the form of spoken utterances, touch-screen gestures, and mechanical button-presses, and presenting output via display on the handset screen, or audio playback. Second, it uses the cellular data network to send those commands to the Promptu server, and to receive the server's output for presentation to the user.
Promptu has built clients for all major mobile programming platforms, including iPhone OS, J2ME, BREW and Android, and deployed products on all major handset brands, including Apple, Nokia, Samsung, Motorola, Sony-Ericsson, BlackBerry, HTC and LG.
Cellular Data Network The cellular data network provides high-speed, bi-directional, digital connectivity between the client and the server. This network operates parallel to, and independently of, the familiar cellular audio network.
The client uses the data network to establish a connection to the server only when it is needed, running conventional Internet protocols to send and receive digital data, including digitized voice. No conventional telephone call is involved. This connection is error-controlled: as long as a suitable RF channel exists, data are guaranteed to flow from sender to receiver without loss or corruption. As a result, audio propagated over this pathway is immune to the drop-outs and distortions that can affect a conventional cellular audio connection.
Server The Promptu server is the computational engine of the system, providing such primary functions as speech recognition, transaction enablement, and messaging. The server also provides secondary functions, such as business analytics and billing.
Speech recognition is the automatic transcription of the user's spoken words and phrases into text. Promptu's unique speech recognition technology is described here.
Transaction enablement comprises interaction with third-party servers to provide the product or function that is enabled by the particular client. Typical examples include reserving and selling train tickets, or dispatching a taxi to a spoken street address.
Messaging is the ability to send and receive text messages on the user's behalf. Notably this includes SMS messages, Twitter "tweets," and Facebook status updates.
Reliability and Scalability
Promptu stipulates to telephone-company SLAs for system reliability and availability-a must-have feature, since Promptu's customers include some of the world's largest telephone and communications companies.
To meet these standards, Promptu servers are composed of diskless compute blades, which boot from an external NAS, and include redundant instances of all key hardware components and external interfaces. Promptu's proprietary Failover Management System (FMS) provides self-monitoring, alerting and self-repair functions for each server.
The Promptu architecture scales smoothly to meet business requirements. An individual server can be expanded by adding more compute hardware. The additional capacity is automatically detected and integrated into the system, with zero down-time. For installations that exceed the capacity of an individual server, multiple servers can be supported by the addition of a front-end load balancer. This same architecture can support geographic redundancy, where required.
Security
Promptu's software architecture policy requires that only kernel processes may run as root, and that every server include a software firewall. All sensitive information sent over the air is transported using HTTPS, with 128-bit encryption (MD5 by default), and all speech recognition is completely automatic, with no human postprocessing or correction. Promptu's security certificates are provided by Verisign ®. To ensure the physical security of the machine, and to provide reliable power, environmental support and Internet connectivity, Promptu servers are located within professionally-run co-location facilities, with human and biometric access control.