Voicescape's Technology

Voicescape's experience in designing, developing and hosting intelligent voice solutions unlocks the benefits of both cost savings and improved efficiencies provided by the latest speech technologies. Voice technology is now proven and when deployed effectively recognition rates can be as high as 96% of that achieved by a human voice. Issues that have historically caused problems, such as the speed of speech, background noise interference, differences in regional accents and intonation, have now been addressed with the latest ASR engines.

A Robust Technology Platform

Voicescape's technology platform is based across two geographically distributed sites, which mirror each other and are of a similar size, capability and resilience. Both sites are integral to our network, and each is self-sufficient in terms of required resources. Management is provided by a Dell supplied EMS subsystem, which also has a distributed component, as well as a centralised overall monitoring and management capability.

The distributed architecture uses a Media Server as the core component, with this server and the other key components replicated at both geographical sites, and managed using the EMS overlay structure.

The architecture is driven by a number of key design requirements:

• Support for shared speech recognition and text-to-speech (TTS) resources
• 'Just-in-time' delivery of capacity
• Evolution of hardware infrastructure in accordance with leading edge commodity class hardware improvements
• High-availability and resilience
• Remote manageability
• Aggregation of physical resources into high-capacity virtual components.

At the core of the Media Server is our VoiceXML Media Server platform. Our VoiceXML platform integrates with leading speech recognition engine vendors, including Scansoft and Telisma which both offer rich speech synthesis support. The Media Server can process PSTN or VoIP media streams and can use speech engines in 'on-board' or distributed configurations. This solution includes:

• VoIP/PSTN Servers
• Speech Recognition Servers
• TTS Servers
• Shared Local Storage Servers
• Per-site Cluster Management Platforms (local component of EMS).

Each of the sites has a homogeneous structure; however, they scale independently. Consequently, adding port capacity is as simple as adding platforms, as most of the Media Server architecture scales linearly and horizontally. The EMS overlay is structured in a hierarchy and servers can be used for any purpose within the architecture - VoIP/PSTN Call Processing, ASR, or TTS resources. The platform architecture is based on commodity class hardware and software, which builds upon the capabilities of our management tools, to provide a hardware platform capable of scaling to support any requirement.