What's New in Speech Server 2004 R2

What's New in Speech Server 2004 R2

Microsoft Speech Server 2004 Release 2 (MSS R2) offers support for creating speech applications in languages other than English (United States) and an enhanced customer experience through improved performance, better reliability, and enhanced error reporting. This topic briefly describes these enhancements and new features.

• Support for New Languages

• Improved Reliability

• Faster Loading of Large Grammars

• Single Computer Configuration for Enterprise Edition

• Ability to Play Audio on Windows Media Player

• More Detailed Error Reporting

Support for New Languages

MSS R2 provides language support for Spanish (United States) and French (Canada) in addition to legacy support for English (United States). The two language packs are installed separately, and each installation includes all the components that MSS R2 and the Microsoft Speech Application SDK Version 1.1 (SASDK) need to support the language engines and development tools.

The new language support provides speech recognition (SR) engines and text-to-speech (TTS) engines, so developers can build applications that contain simple menu branches. For example, an application can prompt, "Press one for English, press two for Spanish."

A new Speech Application Project Wizard, installed with the SASDK, simplifies the development process.

New language support is also built into the MSS R2 analysis tools. You can:

• Filter queries by language of call

• Parameterize reports by language using Speech Application Reports

• Select calls (and other components, such as QAs) by language in Call Viewer

• Filter on event types during database import so that you can better manage the content and the size of your Call Viewer databases

In the Event log files (.etl files), the Message property of each QASummary event will contain the language of the current page. Also, a Language property has been added to calls. This property is represented as a new column in the Call Summary pane and also appears in the Where [property] list.

Also, MSS R2 takes advantage of an upgraded speech recognition (SR) engine, providing better performance by using less memory and a higher degree of speech recognition than its predecessor.

Improved Reliability

Failover is improved in MSS R2 through the use of asynchronous interprocess communication calls between processes such as Lobby, Speech Engine Services (SES), and SES workers.

Implementing asynchronous calls enables SES to recover automatically from error conditions such as remoting call loss and process deadlock, resulting in an improved end user experience.

MSS R2 also expands the use of process recycles. During a process recycle, the responsibility for processing calls shifts from one process to another, typically because the first process has reached its memory limit. The new process takes new calls and the old process finishes processing existing calls.

With MSS R2, in addition to managing memory limit issues, process recycles handle engine failures, engine deadlocks, unhandled exceptions, and timeouts.

Faster Loading of Large Grammars

When a speech application uses large grammars in SES, loading time and memory usage can affect both performance and scalability. In MSS R2, the SR engine enables an engine instance to save its grammar loading state onto disk. The saved state can then be loaded into other SR engine instances to expedite the grammar loading and reduce overall memory usage.

Single Computer Configuration for Enterprise Edition

MSS R2 Enterprise Edition can be configured with TAS and SES on the same computer, enabling you to exceed the 24-channel per processor limit on a single computer configuration. Keep in mind that your hardware capacity is primarily limited by the size of your grammars and applications. But if, for example, you want MSS to handle 30-40 DTMF calls on a single computer, you can now meet that requirement using the Enterprise Edition single computer configuration.

Ability to Play Audio on Windows Media Player

In Call Viewer, you can now use the Play Audio feature to play audio back through Windows Media Player (or another audio player), giving you greater control over audio listening.

More Detailed Error Reporting

When an error occurs, SES provides improved reporting, enabling developers and administrators to diagnose problems and minimize the impact to their system and schedule. This detailed reporting is targeted at the following conditions: