April 11, 2001 Figures of speech
Voice-application activity on the rise
Many software and hardware vendors have recently come to realize that people are more than willing to speak up to get what they really want out of their wireless devices.
Speech recognition is rapidly coming to be seen as a killer application for cell phones, PDAs, Web tablets, and other wireless devices. The technology has been making major headway in the office arena, as businesses adopt automated telephone systems. These systems let callers access company directories by merely speaking the name of the person with whom they wish to talk. The enabling technologies are primarily based on VoiceXML, a standardized language for creating telephony voice-user interfaces that greatly simplify the development of Web-based speech-recognition applications.
According to Cahners In-Stat Group, speech recognition is poised to enter every walk of life as new applications are developed and prices drop. The result is soon to be a speech-recognition-enabled world where the technology can be found at every office, call center, wireless provider, and e-vendor.
"The explosion of this market is due to several factors," says Brian Strachman, senior analyst with In-Stat's Voice Applications Service. "First, microprocessor technology is advancing to the point where speech recognition products are made faster, better, and cheaper. Secondly, demand from service providers and voice portals are creating new markets for speech-recognition products. Finally, development tools such as VoiceXML make it easier to create cutting edge products."
"Coupled with the explosive growth in contact centers and ever increasing demand for CRM [customer relationship management] applications, there is great opportunity for both software providers and application developers to capitalize on the speech-recognition explosion."
Cahners In-Stat presents its findings in the first ever comprehensive speech-recognition market report, "Speech Recognition: Listen to the Market Explode." The Report analyzes the technology and its future, segments the potential markets and forecasts their potential, and provides profiles, market shares, and forecasts. In addition, the report provides forecasts for the speech-recognition software market through 2005, and includes market shares for leading vendors.
In the final analysis, In-Stat finds that, by 2005, annual sales of speech-recognition software engines, the basis for all speech-recognition products, will reach $2.7 billion.
One vendor, Sensory, has just released its Voice Extreme Toolkit, a suite of development hardware and graphical programming tools that allows programmers to turn out demos, prototypes, or entire speech-enabled applications in just a few hours. Available for $129, the Toolkit is a hardware-development platform that includes a Windows-based integrated development environment and works in tandem with Sensory's Voice Extreme IC, a programmable version of the company's RSC-364 speech-recognition chip.
"Unlike traditional speech-recognition development technologies, which require developers to be experts in assembly language, our VE Toolkit is instantly usable by nearly anyone with C programming experience," says Todd Mozer, Sensory's president and chief executive officer. "Now, everyone, from engineers in large companies who want a quick way to sell speech-enabled concepts to their management, to developers with modest-volume applications, can add speech recognition or speech activation to their designs without high-cost tools or custom printed circuit boards."
Available beginning next month, Sensory's VE Toolkit consists of a main board with built-in speaker, microphone, LEDs, power regulation, switches, and prototyping area, as well as a removable VE module that contains the speech processor and flash memory. Also included: a CD-ROM with development software, sample applications, templates, a manual, help, and more. Once programmed via the included serial cable, the tiny Voice Extreme module can easily be removed and wired directly into demos or products. Using a built-in dynamic memory handler, the Voice Extreme chip uses the flash memory to store the program as well as speech prompts (synthesis), speaker-independent weight sets, speaker-dependent templates, record and playback data, music data, and other information.
At last month's CTIA in Las Vegas, Lobby7 launched an integrated speech/graphics interface engine that is targeted at companies designing and deploying mobile communications devices. The company's "multi-modal" interface engine enables wireless users to navigate through menu-driven systems while simultaneously using a mix of voice and visual controls. Mobile users can ask questions or make requests to their wireless devices and then receive a real-time, graphical response that is displayed right on the screen of their handheld units.
"The ability to recognize the user as a person rather than a device, and to follow the session logic is critical and very difficult to do," says John H Hill, communication software analyst for Salomon Smith Barney. "LOBBY7 delivers an integrated graphical UI and voice portal that can be accessed via WAP phones or connected PDAs or PocketPCs, allowing users to request graphical content using their human voice."
The VoiceXML-compliant speech/graphics interface engine allows users to interact with applications on mobile handheld devices using a unified two-way speech and visual-input interface that accesses information in real-time through the simultaneous reception of information in speech, audio, and graphical formats. According to Lobby7, the interface can integrate with all major speech-recognition and text-to-speech systems. In addition, the engine employs J2EE and XML-based architecture that is scalable to accommodate the needs of various content sources and users.
SpeechWorks International and VoiceGenie Technologies are also getting into the act through their recently announced partnership in the development of a scalable platform for standards-based speech-driven applications. Set for release in Q3, the forthcoming SpeechGenie system will enable a company's customers to access Web information, conduct online transactions, and manage personal communications through voice commands over cell phones as well as their landline counterparts.
Based on VoiceXML, the SpeechGenie platform is expected to feature pre-packaged, VoiceXML-based application building blocks as well as text-to-speech engine and SpeechSecure authentication, which will enable users to direct their calls, obtain information, and complete transactions automatically, simply by speaking naturally over any phone, anytime.
VoiceGenie also recently introduced VoiceGenie VoiceXML Gateway 5.0, which enables a company's customers or employees to access Web information, conduct on-line transactions, and manage personal communications, including e-mail and voice-activated dialing, by simply issuing voice commands over any telephone. Gateway 5.0 also provides voice-over-IP (VoIP) support. The result, says the company, is a major reduction in line charges, network costs, and IT expenses.
"VoIP capability is especially beneficial to mid-size enterprises and next-gen telcos, as a large proportion of their communications budgets goes to long-distance services," claims Stuart Berkowitz, president and CEO at VoiceGenie. "By utilizing IP telephony for their new Voice Web services, VoiceGenie customers can leverage their data networks by running converged data, voice, and other applications over a single infrastructure."
By avoiding the PSTN, interoffice telephone calls can travel free of toll charges. Voice traffic instead travels over a company's data lines and existing infrastructure at no added cost. VoiceXML Gateway 5.0 is scheduled to begin shipping in Q3.
General Magic recently launched its own VoiceXML-based gateway. Featuring a scaleable, VoiceXML-compliant architecture, the magicTalk Voice Gateway consists of server-based software that integrates VoiceXML, telephony, media, and speech-recognition services. In addition, the magicTalk Voice Gateway can be integrated with any Automated Speech Recognition (ASR) and Text-To-Speech (TTS) software. The software product also includes a Java-based VoiceXML custom extension API (application program interface) that will enable customers to add their own proprietary tags to extend the VoiceXML standard as well as a VoiceXML debugger for the line debugging of VoiceXML scripts.
At CTIA last month, Philips Speech Processing rolled out its Speech-enabled Highway, which utilizes core speech-recognition technologies such as voice portals, speech-enabled directory assistance, and voice-activated voicemail. All of these features can be deployed in both fixed and mobile networks. Philips is also teeming with Intel company Dialogic to speed development of voice-portal and speech-recognition applications within the landline environment.
Philips Speech Processing's "say-what-you-mean" SpeechPearl speech-recognition engine can be integrated into third-party voice-processing platforms so that callers can speak naturally and be understood. SpeechPearl even incorporates a "barge-in" feature that allows callers to interrupt and begin speaking before the automated prompt is finished. The host-based software-only SpeechPearl engine supports 200,000-word vocabularies and 22 languages, while allowing for dynamic switching between vocabularies and languages. In addition, the platform's add-word feature enables application developers to provide for the critical needs of auto-attendant applications.
