Saturday, April 25, 2009

Interactive Voice Response- Speech-based system: Microsoft's TellMe Voice

Microsoft's tellme service


And a demo of TellMe Voice


jp

1 comment:

  1. Speech-based interfaces must enhance a user's experience. The decision to interject key pad response, voice response and live agent interaction, depends on the requirements and will lead to good design of interactive voice response systems.

    Speech based IVR is meant to handle millions of calls and to relieve live agents of the large volume of calls. At the same time, it should provide the required information to the user, minimizing the interaction time. Lufthansa relies on key input and in two steps provides the answer to the customer, making it more efficient than the speech enabled system of Air Tran. In speech-enabled IVR systems, the touch tone key pad or live agent must be availabel for fall back in case of error.

    Interactive speech-enabled IVR systems still rely on speech recognition based on directed dialog through question and answer sesssions. Natural language processing systems using Artificial Intelligence is still in the R&D stage. The speech recognizer is able to pick up a single answer per question. Speech recognition along ith VoiceXML and data integration provide the technology behind speech-enabled IVR systems. VoiceXML is now the defacto standard for voice-enabled IVR systems, though a competing standard SALT (Speech Application Tags) was tried out by the SALT Forum founded in 2001 by Microsoft, CISCO, Comverse, Intel, Philips and ScanSoft. VoiceXML was developed by W3C. Microsoft Speech Server 2007 supports VoiceXML 2.0 and 2.1 in addition to SALT. Major vendors of speech-based IVRs are Aspect, Avaya, Cisco, Convergys/Intervoice, Genesys, Holly Connects, Nortel, Syntellect/Envox, Voxeo, and so on..

    Language support is also an important aspect of speech recognition systems used in IVR platforms. For example Nuance Communications' speech recognition engine (in Microsoft's .NET framework) was chosen by Dubai Airport and Dubai Financial Market, as it provides Arabic language speech recognition and interaction. The language support would require dealing with various dialects, vocabulary and accent differences of a particular language.

    -Joseph Ponnoly

    ReplyDelete