Nuance 8.0 Delivers Lower Costs for Enterprises and Service Providers, Boosts Customer Satisfaction with Superior Accuracy and Personalized Caller Interactions.

Business Editors/High-Tech Writers

MENLO PARK, Calif.--(BUSINESS WIRE)--Jan. 14, 2002

Listen & Learn(TM), AccuBurst(TM) and other advanced capabilities

drive new levels of return on investment for global companies

Nuance (Nasdaq:NUAN), the global leader in software for a voice-driven world, today unveiled Nuance 8.0, a new version of the company's flagship speech recognition and natural language understanding server software. Nuance 8.0 introduces Listen & Learn and AccuBurst, capabilities that enable speech systems to automatically adapt to callers' unique speech characteristics and usage patterns to boost recognition accuracy, optimize user interfaces for the individual and improve utilization of system resources. End users will benefit from faster, more efficient and satisfying phone-based interactions, while carriers and enterprises will be able to offer more cost-effective voice-driven services that increase customer satisfaction and accelerate return on investment.

"With Nuance 8.0, Nuance continues to deliver innovative speech recognition software that leads the industry in accuracy," said Ron Croen, president and CEO of Nuance. "This directly translates into greater usability, higher customer satisfaction, reduced costs and a faster return on investment for carriers and enterprises."

Additionally, Nuance 8.0 provides accuracy improvements for calls placed from wireless and noisy environments; support for the VoiceXML 2.0 standard; and new tools that facilitate partner deployments of systems built on Nuance's revolutionary Say Anything(TM) technology. Nuance 8.0 is supported by a broad range of interactive voice response (IVR) platform providers, system integrators, application developers, telephony hardware companies and voice application service providers around the world including: Cisco, Datria, Edify, Gold Systems, Intel, InterVoice-Brite, NMS Communications, SAP, Syntellect, Telera, VeCommerce, Voxeo, West Corp. and many more.

Listen & Learn Improves Accuracy and Personalizes the Caller Experience

Nuance 8.0's new Listen & Learn features enable speech recognition systems to automatically detect and adapt to the unique speech characteristics of a carrier's or enterprise's broad customer base as well as individual callers. With Listen & Learn, carriers and enterprises can improve accuracy as well as provide each caller with a tailored experience. Listen & Learn features include Auto-tuning, Personalization Kit(TM) and Dynamic Language Detection(TM).

Auto-tuning

With Auto-tuning, speech recognition systems can be automatically tuned to account for characteristics common to a specific carrier or enterprise's caller population. For instance, a carrier serving subscribers in the southern United States can use Auto-tuning to automatically adapt its voice-driven customer care system to account for the unique southern accent common among its subscribers. Auto-tuning also enables speech recognition systems to automatically detect phrases and words that callers often use when interacting with the system, improving the system's ability to accurately recognize these frequently spoken expressions. In many cases, Auto-tuning can reduce error rates by as much as 60 percent resulting in faster customer transactions and improved quality of service. Moreover, because Auto-tuning helps automate the tuning process, development effort, time and costs can be reduced, freeing valuable IT resources to focus on other projects.

Personalization Kit

Personalization Kit provides patent-pending technology that analyzes voice characteristics and environmental conditions of an individual call in real-time. Personalization Kit can determine whether the user is a man or woman, identify the type of phone (wireless or landline) and assess background noise levels present during a call. Enterprises and carriers can use this information to create user interfaces tailored to each caller, making services easier to use. For instance, a voice recognition system may have difficulty hearing a very soft-spoken person calling from a noisy, public location. Using the Personalization Kit, the system can prompt the caller to "please speak a little louder" so that the individual may be heard. Increased revenue-generating opportunities are also possible by offering callers real-time, targeted marketing information on products and services based on demographic characteristics, such as gender.

Dynamic Language Detection

Dynamic Language Detection allows voice applications to automatically detect in real-time the language the caller is speaking. The caller simply begins speaking in his/her preferred language and the system understands and can interact with the caller accordingly. The system can also recognize with high accuracy if the caller mixes languages or speaks in a heavy accent periodically throughout the call. This is particularly beneficial when offering voice-driven services to multi-lingual markets such as English-French speaking Canada and many English-Spanish speaking markets throughout the United States.

"Nuance 8.0 speech recognition software has a valuable set of features and capabilities such as Listen & Learn, Personalization Kit and Dynamic Language Detection that help corporate enterprises address their most important business goals such as improved customer service, customer retention and reduced costs," said Joseph G. Brown, president and chief executive officer of Edify. "By integrating products like Nuance 8.0 into the Edify Electronic Workforce(R) vCSR(TM) platform, we can continue to offer our customers the latest speech solutions that are targeted to meet their specific objectives."

AccuBurst Leverages Under-utilized Hardware to Improve Accuracy

Breakthrough advances in Nuance's distributed architecture have produced AccuBurst, a patent-pending feature that dynamically improves accuracy and optimizes system CPU and memory resources by detecting the number of callers interacting with a speech recognition system at any given time. During brief periods of peak call volume, a speech recognition system's CPU and memory resources are utilized near capacity. However, during off-peak conditions, speech recognition systems use only a fraction of available computing power, with remaining CPU and memory resources sitting idle. Unique to Nuance 8.0 and the speech recognition market, AccuBurst can automatically detect low call volume and dynamically apply the idle computing power towards improving accuracy, reducing recognition error rates by as much as 30 percent. Idle "backup" computing power that is often available to provide system redundancy can also be leveraged. This not only improves the quality of service experienced by customers, but also lowers costs through more efficient use of computing hardware. AccuBurst is ideally suited for enterprises, carriers and hosting companies that face very large daily or seasonal peaks in call volume.

"High recognition accuracy is vital if callers are to complete transactions on speech systems without the involvement of live agents, yet still with high levels of satisfaction," said Dan Hawkins, voice business program manager of Datamonitor. "Innovative features such as AccuBurst can help carriers and enterprises increase their accuracy and therefore increase service levels. They can also improve automation rates and help achieve the cost savings this implies."

Improved Performance Lowers Hardware Costs

In addition to AccuBurst, Nuance added a number of other improvements to its core software architecture that further improve overall CPU and memory utilization. For instance, memory requirements are reduced by up to 75 percent for North American English applications, and CPU requirements are reduced by up to 50 percent for deployments running on the Sparc Solaris operating system. These improvements help carriers and corporations reduce their overall hardware requirements and dramatically cut the cost of ownership associated with a voice-driven system.

Support for VoiceXML 2.0 Eases Deployment Effort and Lowers Costs

Nuance 8.0 provides support for speech deployments built using VoiceXML 2.0 -- the latest version of the industry standard markup language for voice applications. Using VoiceXML 2.0 with Nuance 8.0 speech recognition software, carriers and enterprises can leverage their existing expertise and investment in Web infrastructure to reduce the time, effort and cost associated with deploying voice-driven services. Partners offering VoiceXML 2.0 based solutions built on Nuance 8.0 can take advantage of the most recent release of GrXML, the VoiceXML standard's grammar format specification. Furthermore, advances in Nuance 8.0's architecture enable faster and more efficient grammar processing for VoiceXML applications. These new features are critical to supporting the more than 70 companies in 24 countries who are currently using Nuance to power their VoiceXML solutions.

Say Anything Enables Natural, Free-Style Speech Interaction

Nuance 8.0 offers partners newly available tools for deploying and customizing applications built with Say Anything technology. With Say Anything, users can speak more freely and naturally when interacting with speech-enabled systems. Applications can greet callers with open-ended prompts like "what would you like to do?" and users can respond with wide ranging answers such as, "I have a problem with my last bill and I need some help," or, "I'd like to, um, get my, um, bank balance." Leading carriers and enterprises are already replacing their touch-tone call routing systems with Say Anything applications resulting in simplified, easy-to-use voice-driven services that provide faster access to information, reduce reliance on expensive live agents and improve customer service. To date, over 4,000 ports of Say Anything applications have been deployed.

Nuance Continues its Leadership in Accuracy for Noisy, Wireless and Hands-free Environments

With the rapidly growing number of cell phone users around the world, carriers and enterprises need to ensure that their voice-driven services can handle the many challenges associated with calls placed using wireless devices. These challenges often include background noise resulting from usage in public locations (e.g., crowded airports and restaurants or a busy street) or while driving (e.g., wind noise, radios, horns). Nuance 8.0 rises to the challenge by reducing error rates by over 10 percent for calls placed from noisy, wireless environments. The software also filters out "side speech" -- speech from others in close proximity to the caller whose words are not intended for the recognition system. Nuance 8.0's superior performance for wireless and noisy conditions, coupled with other accuracy enhancing capabilities such as Listen & Learn and AccuBurst, enable carriers and enterprises to deliver the market's most accurate voice-driven services.

Additional New Features and Improvements in Nuance 8.0

Nuance 8.0 supports new operating systems and various other performance improvements to further increase customer service and reduce deployment cost:

-- Alphanumeric and name recognition is improved, providing up to a 50 percent
reduction in error rates. These improvements are made possible when used in
conjunction with Nuance's voice authentication software, Verifier 3.0.

-- Support for Linux and Windows 2000 lowers deployment costs for carriers and
enterprises that have already deployed these operating systems throughout their
infrastructure.

Pricing and Availability

Nuance 8.0 is available now in 26 languages for global deployments. Nuance 8.0 for Windows 2000 and Solaris can be downloaded via the Nuance Developer Network at extranet.nuance.com. Nuance 8.0 for Linux and AIX will be generally available in Q1 2002. Pricing begins at $200/port.

About Nuance

Nuance's speech recognition, voice authentication and text-to-speech software is changing the way companies interact with customers over the phone. In markets around the world, leading carriers and corporations -- including British Airways, Nomura Securities, OnStar, Sprint PCS, United Parcel Service, Vodafone, and many more -- use Nuance's software to reduce costs, increase customer satisfaction and retention, create new sources of revenue and improve security. Nuance's software powers a range of innovative applications including automated directory assistance, voice-activated dialing, self-service banking and voice-driven customer relationship management. The company is headquartered in Menlo Park and has offices around the world. For more information, visit www.nuance.com or call 1-888-NUANCE-8.

This press release contains forward-looking statements, including, for example, those relating to the development and availability of new products, including for example, the general availability of Nuance 8.0, the applications that may use Nuance 8.0 and the benefits that may be achieved by Nuance 8.0 customers and end users. There is no assurance that the results contemplated by any forward-looking statements will be realized. The following factors, risks and uncertainties, among others, could cause actual results to differ materially from those described or implied in this press release's forward-looking statements: the uncertainty of Nuance's ability to successfully develop and make generally available new products, including, for example, Nuance 8.0, the risk that customers might not deploy expected applications, or any applications, that run with Nuance 8.0; and other factors described in Nuance's filings with the Securities and Exchange Commission, including but not limited to Nuance's last-filed prospectus and quarterly report on form 10-Q. Nuance does not undertake to update any oral or written forward-looking statements that may be made by or on behalf of Nuance.

Note to Editors: Nuance, AccuBurst, Listen and Learn and Dynamic Language Detection are registered trademarks or trademarks of Nuance Communications, Inc. Intel is a trademark or registered trademark of Intel Corp. or its subsidiaries in the United States and other countries. All other trademarks are property of their respective owners.