Speech gains momentum

Adoption of speech technology could be encouraged by consolidation after IBM last week gave its backing to the merger of voice software firms ScanSoft and SpeechWorks.

Last month, ScanSoft said it would acquire SpeechWorks for about $132m, creating a powerful combination across speech recognition, text-to-speech, speaker verification and dictation. Last week, the merger gained greater traction when ScanSoft announced an agreement with speech pioneer IBM that will see the firms work together on speech-enabling and cross-selling applications.

ScanSoft plans to port its telephony applications to IBM's WebSphere server software; also, ScanSoft's VoiceRequest and DirectoryAssistance applications will be compliant with the VoiceXML 2.0 standard and interoperable with IBM's WebSphere Voice Application Access portal. IBM and ScanSoft will also sell each other's products. The agreement adds to ongoing consolidation that has seen ScanSoft acquire the speech units of Lernout & Hauspie and Philips in the past two years.

In a research note, analyst firm Gartner said the ScanSoft-SpeechWorks deal could help the fledgling speech market by combining ScanSoft's international reach with the R&D of SpeechWorks.

"Each will be better positioned to compete in a consolidating speech market where value comes more from the delivery of applications and solutions than from core speech technology," Gartner analysts commented.

"SpeechWorks will obtain a needed approach for preconfigured enterprise products while ScanSoft obtains professional services for advanced application integration," Gartner said. "Both companies also have strong partnerships with different areas of Microsoft as well as with the Speech Application Language Tags (Salt) initiative, thus providing common direction for that key alliance. Both companies also have VoiceXML products."

ScanSoft international marketing director Mark Erwich said the effects of the deal will be seen most strongly in telephony and embedded applications such as those for the automotive sector.

"Safety legislation will mean there is a greater need for [in-vehicle] voice capabilities - for example, to provide directions, or turn on the heater or radio," Erwich said.

Chip giant Intel is also pushing speech technology. The firm last week released software under open-source licensing terms that allows developers to build devices that read lips the way humans do, to improve speech comprehension.

Intel's OpenCV computer vision library toolkit and Audio Visual Speech Recognition (AVSR) software use facial movement detection algorithms to aid recognition, particularly where background noise makes orthodox speech recognition difficult.

Have your say: reply to IT Week