Microsoft Corp.s Speech Server 2004 will cause significant changes in the way companies add speech capabilities to enterprise applications. Speech Server 2004s lower software costs and familiar development tools will help transition what has been a black-box approach to speech application development to a more open and extensible method.
Click here to read the full review of Speech Server 2004.
2
Microsoft Corp.s Speech Server 2004 will cause significant changes in the way companies add speech capabilities to enterprise applications. Speech Server 2004s lower software costs and familiar development tools will help transition what has been a black-box approach to speech application development to a more open and extensible method.
The final beta of Speech Server worked well in eWEEK Labs tests. However, we believe service and support for applications will be as critical for this engine as it has been for other platforms. The reason: Many companies biggest problem will be a lack of in-house expertise required to build an easy-to-use IVR (Interactive Voice Response) system.
Microsoft offers a $7,999 Standard Edition and a $17,999 Enterprise Edition; both have a per-processor price scheme. There is a per-port limit for each installed copy: The Standard Edition supports 24 concurrent open ports per server, and the Enterprise Edition supports 96 concurrent ports per node. Both editions will be available next month.
Microsoft makes it relatively easy for corporate developers to code an application using the Speech Server API and Visual Studio .Net, but Speech Servers biggest advantage over competing platforms is that it allows companies to build multimodal applications. For example, a companys customer support Web site can also become an IVR application. And applications can be designed to work more fluidly with multimodal devices, such as Windows Mobile phones.
Speech Server comprises two parts, the Speech Application Developer Kit and the Speech Server deployment components. The application developer kit includes ASP .Net speech controls, plus grammar and prompt tools. In tests, these components provided simple ways for developers to build or speech-enable an application.
Because Speech Server uses Visual Studio. Net, a development team familiar with Visual Studio should be able to pick up knowledge and take the reins with an application even if an outside integrator built the first version.
The deployment components include the SES (Speech Engine Services), TAS (Telephony Application Services), and the TTS (Text-to-Speech) engine and database. Also included is an interpreter for SALT (Speech Application Language Tags) that make speech-enabled data on a Web page accessible through the IVR system as well as via keyboard or mouse interaction.
While Speech Server capably abstracts development from the hardware layer, managing the software layers that create the abstraction is more cumbersome. A more comprehensive tool set would make management easier. For example, Speech Server supports Intel Corp.s Dialogic cards through Intels TIM (Telephony Integration Manager) software, and we could use only Intels software to manage the cards. Thus, when we needed to reset the card, we had to disable the Speech Server application connecting to that card through the MMC (Microsoft Management Console) and then disable the card through TIM.
The MMC manages the SES and TAS components and allowed us to assign connections and manage parameters such as timeouts. The console will be useful in managing connections in Enterprise Edition deployments, where SES and TAS usually run on different servers to form a node. In the Standard Edition, all components run on a single server.
We would also like to be able to view call volume and application flow through a reporting tool. This would help developers determine how customers interact with an application and show where to tweak an application to improve its design.
In the beta, reporting is handled through Logman, the event viewer and performance monitoring tool. In tests, Logman required a lot of manual intervention or scripting to manage and reported data into trend data.
Technical Analyst Michael Caton can be reached at michael_caton@ziffdavis.com; additional testing by Clarence Arthur.
Be sure to add our eWEEK.com developer and Web services news feed to your RSS newsreader or My Yahoo page: