ScanSoft Merges Engines in New Speech-Recognition Release

With the launch of OpenSpeech Recognizer 3.0, ScanSoft is focusing future development on a single speech-recognition engine that combines code from many engines.

ScanSoft Inc. on Tuesday announced its newest speech-recognition engine—a product that, for the first time, combines into a single product the engines the company acquired last year.

ScanSoft, of Peabody, Mass., launched OpenSpeech Recognizer 3.0, the first major update to the platform that it acquired after its merger last year with SpeechWorks International Inc.

Along with adding new features, the release marks the merging of code from the SpeechPearl engine into the OpenSpeech Recognizer product.

Earlier in 2003, ScanSoft acquired SpeechPearl as part of its purchase of the speech-technology business of Royal Philips Electronics.

ScanSoft will continue to support SpeechPearl customers but is no longer planning to develop new versions of SpeechPearl, according to Steve Chambers, president of ScanSofts SpeechWorks division.

"We are going to one [recognition] product with this release of OSR," he said. "It brings together the SpeechWorks and Philips technology."

With OpenSpeech Recognizer 3.0, ScanSoft has boosted the accuracy of speech recognition, reducing error rates by an average of 25 percent, Chambers said. It also has brought the number of languages supported up to 44.

The new version also focuses on incorporating more natural-language capabilities through ScanSofts SpeakFreely technology and provides an expanded pronunciation dictionary that covers proper names, thus allowing applications such as auto-attendants to more accurately recognize names.

/zimages/3/28571.gifClick here to read about ScanSoft bolstering its speech call routing.

OpenSpeech Recognizer 3.0, which will be generally available May 31, runs on Microsoft Windows 2000 and 2003 as well as on Red Hat Linux 7.2 and Red Hat Advanced Server 2.1. Support for SPARC Solaris 8 will follow in July, according to ScanSoft officials.

With code bases combined from the companys speech-recognition engines, the speed at which ScanSoft can incorporate new innovations into follow-on releases will increase, Chambers said.

ScanSoft already plans to begin integrating its Dragon NaturallySpeaking dictation technology into the next OpenSpeech Recognizer version, bringing dictation capabilities to network-based speech technology, Chambers said.

It also is looking at ways to add more personalized speech features and at improving interaction between embedded speech technologies in devices such as phones with network-based speech.

The next version, OpenSpeech Recognizer 4.0, is planned for release in early 2005, Chambers said.

In other speech news Tuesday, Envox Worldwide released the next version of its voice-application development platform, which offers expanded VoiceXML, speech, Web services and voice-over-IP (VOIP) capabilities in a single platform.

Envox 6 includes support for the latest speech platforms from ScanSoft and Nuance Communications Inc. as well as Web services support for protocols such as Simple Object Access Protocol, XML and Web Services Description Language to allow integration between voice applications and enterprise applications, the company announced.

/zimages/3/28571.gifCheck out eWEEK.coms Enterprise Applications Center at http://enterpriseapps.eweek.com for the latest news, reviews, analysis and opinion about productivity and business solutions. Be sure to add our eWEEK.com enterprise applications news feed to your RSS newsreader or My Yahoo page: /zimages/3/19420.gif http://us.i1.yimg.com/us.yimg.com/i/us/my/addtomyyahoo2.gif