Creating next-gen speech applications with open source building blocks

Businesses are choosing to interact with their customers using speech based, natural language interfaces that are friendly as well as efficient. Advancements in speech processing technology have made it possible to create systems that can 'talk' to customers and enable transactions in a customer- friendly as well as operationally-efficient manner. One of the key enablers for creating natural and customer focused speech systems is VoiceXML (http://www.voicexml.org). As an open specification for writing voice applications VoiceXML helps businesses break out of lock-ins created by proprietary IVR applications. It also facilitates the use of advanced speech processing technology and provides a customer-focused approach to application development

Obelisk^TM is a software plug-in for Asterisk (www.asterisk.org) - the world's leading open source telephony engine and tool kit - that converts Asterisk into a VoiceXML capable speech application platform. The combination of Asterisk's advanced call control capabilities and Obelisk's support for VoiceXMlL and MRCP standards delivers a highly advanced, robust yet affordable speech platform. Obelisk^TM's unique bolt-on design allows enterprises to deploy voice driven applications that can leverage best-of-breed speech technologies. Furthermore, by supporting industry standard communication protocols Obelisk^TM allows application architects to adopt a build-as-you-go strategy. This means that application designers can add specialized features to their systems as and when needed.

Build-as-you-go with a “Lego block” approach

Obelisk allows application designers to minimize the dependencies in a project’s critical path. The system allows the basic application flow to be built using simple record and play-back functions while other pieces continue to be developed or tuned. Obelisk makes use of the open source media resource control protocol (MRCP) stack from UniMRCP (www.unimrcp.org) which allows third party speech recognition and synthesis systems to be added or replaced in a simple manner.

Key features include

• A modular design that plugs into proven open source speech  components
• Support for major VoiceXML 2.0 elements and attributes
• Ability to connect state of the art speech recognition systems as and when required via MRCP interface
• Support for DTMF and speech grammars
• Support for audio caching
• Easy to install, configure and deploy