Grammar Caching

Speed Up Your Speech Application

(SEPT 2007) — One of the new features in the 8.0 release of the Speech Engine — now available on all platforms — is grammar caching.

This means that once your application has loaded a grammar, the Engine will save a copy of that grammar to disk.

The next time your applications loads that grammar, the Engine will be able to use that cached copy to speed up loading times. For large grammars, this can significantly reduce loading time.

About Grammar Loading and Caching

When the Speech Engine loads a grammar, it usually loads an SRGS grammar file from disk (either in XML or ABNF format). Internally, the Engine compiles the SRGS grammar into a binary format used by the Engine.

This compilation is usually quite fast, but as grammars get large, it can begin to take a little bit of time. If your application needs to load a large grammar many times throughout the day, even a hundred milliseconds can start to add up.

To help alleviate delays, the 8.0 release of the Speech Engine will automatically save a compiled copy of a grammar the first time it is loaded.

The next time an application asks the Engine to load the grammar, the Engine will use its saved copy. Since that cached copy is already compiled, there is no delay while the Engine compiles it.

When the Engine caches a grammar, it generates a unique identifier (a hash) for the loaded grammar, and assigns that hash to the cached copy. Whenever your application loads a grammar, the Engine checks the hash of the requested grammar against all the cached grammars and will use a cached copy if it finds a match.

Since the hash changes if the grammar changes, the Engine will automatically know to make a new copy of any updated grammars.

Controlling Caching

Inside the Engine installation directory, there is a new directory called config, inside of which you will find a file called server_cache.conf that controls various options for grammar caching.

enable–cache: Controls whether caching is enabled or disabled. If it is set to 0, caching is disabled; if it is set to 1, it is enabled. By default it is enabled.

time–since–lastaccess: So that the Engine does not fill up your disk space with cached grammars, it will delete grammars if they have not been used for the time set by this parameter. Times are specified in HH:MM:SS. By default this parameter is set to 10:00:00 (10 hours).

cache–check–interval: This parameter controls how often the Engine should check to see if any cached grammars have reached the time limit specified by time–since–lastaccess and thus need to be deleted. Times are specified in HH:MM:SS. By default it is set to 00:00:50 (50 seconds).

New Features in Version 8.0

The September release of LumenVox products, including the Speech Engine, Platform, and Tuner, is our 8.0 release that includes the new grammar caching feature.

This release includes several other noteworthy changes, including optimized grammar compilation, improved confidence scores, and a reworking of the older concept/phrase grammar behavior. The Speech Engine help document has a complete list of changes.

For information on downloading the latest releases of the LumenVox software, please contact us. It is a free download for users with current software maintenance packages.

Solaris Speech Recognition

As of 8.0, LumenVox has expanded the platforms for which the Speech Engine is available to include Sun Solaris 10 running on SPARC hardware. If you are a Sun user interested in trying the Engine on your Solaris installation, please contact us for more information.

Aculab Easy Speech Bundle

If you're interested in getting started with LumenVox speech recognition, we have recently announced the "Easy Speech" bundle with Aculab.

This bundle, available at a special starter price for a limited time, has everything you need to quickly build and run speech–enabled applications using VoiceXML.

The bundle includes four LumenVox Speech Engine Lite licenses, a Prosody X media processing card (with one E1/T1 trunk and support for up to 150 VoIP channels), and four licenses for the Aculab VoiceXML/CCXML interpreter.

Join Us At Astricon

We will be exhibiting at Astricon, Sept. 24–28 in Phoenix, AZ.

If you are interested in getting started working with speech recognition on the Asterisk PBX, Stephen Keller will be presenting a tutorial called Designing Your First Asterisk Speech Application at 11 a.m. on Sept. 26.

For those who haven't yet registered, you may use our special discount code when registering: AC–US–07–LUMENVOX–9195. More >

New Training Videos

New speech recognition videos are being added to LumenVox.com all the time. Recent topics include:

These are just the latest videos. Our archive includes more than 30 videos on a diverse range of subjects. More

© 2016 LumenVox, LLC. All rights reserved.