Join us for a presentation on the results of the 2009 Alignment Index for speech self-service and a discussion among a panel of speech experts on what the results mean for the speech industry and consumers. Dimension Data, Cisco and Tellme have joined forces with this initiative and will take attendees through the results of the survey, and then the panel will join in to discuss areas of misalignment between the industry and consumers and to give recommendations for how to address these areas going forward.

It is not practical to automate 100 percent of your call volume with a self-service voice portal solution. So how do you get to the delicate balance between automation and customer satisfaction? In this session, you'll see the optimum balance between self-service and agent assistance to maximize your investments, enhance your customer service and drive down your contact center costs while weighing the pros and cons of CSAT versus self-service or agent-assisted calls.

What clever services are emerging that combine human and machine speech recognition? How can these services deliver amazing results as reasonable costs, including voicemail-to-email conversion, sophisticated self-help systems, and high-reliability services? Could recognition become so advanced that dialogue technology is the limiting factor? This session provides an insider's view into the technology and economics behind emerging speech recognition services, including hidden humans in a system, where these services are today, and how you can get them.

Audio is becoming commonplace in legal and regulatory investigations, but recordings that are stored outside of an information governance strategy can result in high costs and risks to the organization. This session will discuss the growing need to automate audio governance processes, including real-time monitoring, retention management, and legal hold. Additionally, you will be introduced to meaning-based computing, an automated approach that enables governance by processing audio information based on its conceptual and contextual meaning.

There is plenty of debate around the topic of the legal standing of voice biometrics in the court of law in the U.S. and Canada. This session will highlight various legal cases with precedence, examine compliance across industries, and discuss the issue of privacy as it relates to protection of personally identifiable information. Examples of deployments in the insurance, financial, and healthcare industry will be presented.

In today's fast-paced, mobile world, businesses interact with customers in a variety of ways. It's not just your website, agents, and IVR anymore. Nowadays, customer touch points can include emails, texts, and podcasts. And don't forget the more traditional contact methods through advertisements, commercials, and, of course, face-to-face in-store interactions. What's a company to do with its speech solution in such a diverse environment? This session will explore how your solution should fit into your overall customer experience strategy.

When your customers call your IVR system, they're often holding their bills in their hands. Customers use their bills, but bills aren't organized to provide the information they need to navigate the call. Even something as simple as an improperly formatted account number can cause trouble. Does your bill help your callers through the call, or does your bill persuade them to talk to a live operator instead? We'll discuss how to align your bill and your IVR system to help reduce expensive human-handled calls.

Increasingly, callers will be reaching contact centers through speech-enabled directory assistance and voice search. Callers will thus be more accepting of speech automation and expect a rapid response. Volumes of calls will increase. This "Voice Web" will demand "voice sites" and require a change in the operation and philosophy of call centers. Analyst Bill Meisel will give practical advice on dealing with this paradigm change.

Are we witnesses to the dawn of a new interaction age and the apex of IVR customer service? What major shifts will happen? How must businesses and contact centers adapt? What part will speech play? Factors such as increased appreciation for customer experience, voice in the data channel, P2P social models, handheld computers, and the distribution of information are driving upheavals in customer service, from product ordering to tutorials, including contextual mobile interactions that demand greater power and flexibility in the machines we interact with. Get ready for dramatic changes now.

The multimodal experience brings both new excitement and new challenges. Error handling normally found in a VUI or GUI alone doesn't always make sense or work well when combined into multimodal design. This session explores several challenges to consider when designing error-handling strategies for a multimodal experience. Looking at a multimodal prototype for a mobile device, design strategies, and usability results from the prototype will be presented to help illustrate these error-handling design considerations.

In most systems, the strategy for dealing with callers who are struggling is uniform regardless of their state in the system. As a result, some callers become frustrated because they are not transferred to a human agent soon enough and call duration is increased needlessly. A new strategy has been developed based on a predictive model that takes into account the probability of completing a task given the history of the caller's previous interactions with the system. Incorporating such predictive models into live systems to help struggling callers can have a positive impact on caller satisfaction and average call duration.

This presentation will explore how to get more from your utterance collection by examining audio files using speech visualization software. In a typical tuning exercise, it is common to look for out of grammar utterances and misrecognized words, but sometimes it is also important to be sure some of these problems are not caused by other issues related to the quality of the audio or improper setting of various parameters. This presentation will show ways to overcome these problems by displaying waveforms on-screen and pointing out various acoustic phenomena.

While the keywords and phrases selected by designers may mesh well with the background information gathered during the discovery process, in many cases they may be suboptimal for speech recognition. This session will involve a high-level discussion of phonetics and acoustics designed to provide prompt writers with an understanding of the impact the phrases they choose to use in their prompts (and their grammars) have on speech recognition and ultimately on the caller experience.

You've gone through design and implementation and have just deployed that new application. What should you expect? How will you know if the application is meeting your objectives? What if your callers are not ending up in the right place. What if they are ... lost? Join us for a practical look at how you can equip your callers with a package of "bread crumbs" they can follow. Learn best practices and simple analysis techniques to optimize any application.

Nobody likes a long, congested commute, and customers don't like long, misguided calls in an IVR. Many times callers abandon the IVR for reasons that do not relate to the exact module in which they hung up. The Caller Traffic Light is a high-level flow that exposes these trends by displaying cause and effect. Caller Traffic Lights not only express caller satisfaction in a simple graphic flow chart, they provide an invaluable reference tool on measured customer behavioral patterns.

Usability and tuning are essential to successful VUI design and caller satisfaction. Together, they provide more benefit to an application than either would alone. Traditional approaches separate these two processes into distinct development phases, but recognizing the overlap in purpose of the two fosters a deeply collaborative approach, where information is fed from usability engineers to tuning scientists and vice versa. We will provide real-world examples of this synergy between tuning and usability.

Customer focus groups are an effective method for gathering information about what people think and how they feel about a current application, service, or product. WOZ testing is an effective tool as well for zeroing in on the usability of specific, future-focused speech applications planned for deployment. Yet, conducted in isolation from one another, these tests can create ambiguous and/or confusing conclusions on key next steps to take when refining your application designs. By combining customer focus groups with WOZ/usability testing techniques in a unique hybrid approach, the speech application/IVR is treated as a consumer product that is both discussed in general terms (the current design) and directly experimented with (new design alternatives) in the same session.

A recurring speech application challenge is around the recognition of alphanumeric patterns. Such identifiers are often an integral part of a company's business operations, and capturing such identifiers from a caller is often crucial to what functionality can be implemented in a speech application. The author will present a number of strategies that demonstrate that the solution to such challenges can only be achieved with a tightly integrated combination of VUI design, speech science, and application logic. Data from two applications will be presented to show how it was possible to substantially increase the success rate for challenging alphanumeric recognition tasks over several tuning cycles, analyzing both call recordings and transcriptions by implementing a combination of strategies.

Effectively recognizing alphanumerics is a difficult task, even for humans under optimal telephony conditions. For an automatic speech recognizer, the task is harder still. By making the most of the capabilities of SRGS and SISR, combined with some intelligent post-processing, you can achieve good success rates for even this demanding task. Attendees will learn how to make the most of the logic available to them in SISR and use this information to perform database lookups or other secondary matching strategies to improve alphanumeric recognition.

Recognizing letters of the alphabet is a hard problem for humans and machines, but sometimes it just has to be done. There are a number of tips and tricks you can use to make sure your speech application scores an "A as in apple" rather than an "F as in Frank." This practical guide reviews various strategies you can use to make the grade.

The new W3C standard EMMA 1.0 allows the representation of extended ASR results. Many new annotations are available to open up exciting new scenarios for voice and multimodal applications. The focus of this presentation is on the availability of not only N-best ASR results, but also lattices which will offer richer ASR results, especially with the use of SLMs for recognition. We will highlight the architectural changes now made possible by the introduction of EMMA 1.0 within VoiceXML platforms, as well as the impact on voice application development.

Spending on service-oriented architecture (SOA) is expected to grow significantly throughout 2009, and be one of the most significant areas of technology investment for the enterprise. New technologies make it possible for contact centers to tap into the SOA to drive more advanced speech applications with shorter development times and less risk. This session will focus on the technical standards and practices that allow a speech-enabled IVR to leverage the benefits of an existing or upcoming SOA investment with a heavy emphasis on practical how-to's and real-world examples drawn from the experiences of two large financial services companies.

Phonetic-based real-time monitoring systems aim to provide instant awareness of call center interactions. This session will discuss the state of the art in implementation and deployment of real-time monitoring solutions, including integration with VoIP telephony platforms and customization of triggers and actions, and will present a new phonetic-based speech engine optimized for processing thousands of audio streams in real time with sub-second latency, as well as novel approaches to improve search accuracy such as auto-thresholding.

This presentation will discuss the use of speaker recognition techniques for conducting language proficiency testing and rating. The proficiency testing is being used for 60-plus languages, and the speaker verification system introduced is a language-independent and text-independent system that is utilized to eliminate the need for an active proctor in a long-distance oral proficiency examination. We will discuss techniques derived from the speaker recognition discipline that are used for the automatic rating of oral proficiency interviews. The cases presented here are solutions that are currently being utilized in a real system.

This presentation will describe new technology developed by Next IT Corp. that enables the expansion of IVRs to emulate humanlike interactive behavior.

Of all natural language processing (NLP) technologies, automated translation of speech comes the closest to the ideal of removing language barriers. In this presentation, Donald DePalma surveys the state of the technology for machine interpretation, probes the market potential for voice-activated instantaneous interpretation, and discusses the trends and challenges facing the sector.

This presentation describes and demonstrates DICIT-a dialogue management-based natural language multimodal interface for controlling and interacting with a television. Using multi-token freeform requests, users can not only control basic TV functions but also efficiently filter and locate programs of interest from thousands of programs listed in the electronic program guide. DICIT also includes far-field microphone technologies to enable users to interact from a distance-instead of using a close-talking microphone.

Information access is a daunting task when it comes to context-based, unstructured voice search on large collections of free-text documents such as frequently asked questions. Menu-driven approaches to voice search are inefficient and require intensive application design. Learn about a new information-retrieval-based solution that improves usability and streamlines design. You will learn about information-retrieval-based solutions; techniques for retrieval that includes VUI, text processing, and query construction; and research in context-based, unstructured search.

The increasing need for natural voice interfaces to quickly and easily explore the wealth of content on mobile devices is giving rise to new technology challenges in speech and language processing. Among these challenges is the ability to learn and adapt to the user with minimal or no supervision. Learning can involve optimizing models, grammars, and the user interface based on many data streams including audio, usage profile, and meta information. This talk will review our ongoing research toward creating advanced mobile voice search services that adapt and learn with minimal or no supervision. Our research enables adaptation to users' voice and other usage data to provide personalized models and grammars.

Abstract dialogue management concerns what action or response dialogue systems should take in response to user input. Recently, academic researchers have been pursuing methods for automating the design of dialogue management using machine learning techniques such as reinforcement learning. This talk will provide a survey of these techniques from an industry perspective, critically evaluating to what extent they are ready for commercial deployment. This talk will also highlight machine learning approaches with more practical import.

This presentation will discuss the deployment of an outbound automated application called Auto Dialog. AutoDialog helps individuals better understand their medical issues by supplying them with tailored information based on their health history and enables individuals to become more actively involved in their healthcare with their providers. The solution also provides the business benefit of reducing claims costs by focusing on proactive health care and preventative disease management and can reduce the amount of incoming calls for companies by addressing requests for basic information. The presentation will cover business challenges addressed by Health Dialog, implementation methods, and best practices for outbound applications.

PIXSAN and Loquendo have used TTS to develop a speech-enabled Electronic Program Guide for set-top boxes, resulting in improved accessibility for all users, especially users with visual impairments. Williams will demonstrate this new user interface and examine the complex process of speech integration with the on-screen navigation found in today's TV reception.

Because of recent legislation, Boston Sand & Gravel Co. needed a solution to allow mixer truck drivers and customers to interact with their phone system in a safe, hands-free manner. A speech-enabled auto-attendant with customizable VoiceXML templates allows them to retain the IVR functionality customers had grown accustomed to, while adding speech-driven menu options. Callers now speak a prompted menu option or the name of the person or department they wish to reach.

PeopleFind is a cloud-based, speech-enhanced addressbook for connecting people and teams. Integrating ASR and TTS technologies form Loquendo and Nuance, PeopleFind enables and speech-assisted directory for business which unifies contacts across personal, social, and corporate directories.

A multimodal user interface for a call center agent provides a framework for evolving the completion of an entire transaction to the end user. In the call center, voice and graphics can be used interchangeably by the agent in a step-by-step manner to follow the sequence of graphical user interface (GUI) screens designed by a programmer. However, voice modality changes the focus from the GUI screens to the caller's model of the transaction-based flow and enables the agent and caller to interact at the caller's pace and flow, with the agent providing guidance only as necessary. This multimodal approach also applies to a hand-held device for self-service by an end user, where the transaction complexity is simplified with voice input and graphic output.

Combining video with voice to enhance multimedia services is an important development in the next-generation call center. Customers often encounter an IVR system when they call a contact center because standard queries can be satisfied without agent intervention. But video offers the potential to enhance the customer's IVR experience. This presentation will discuss the market trends driving the latest advancements in contact center solutions, as well as the key multimedia standards and technical components behind these advancements.

A unique project started by an energetic government officer in Uttar Pradesh, India, has become an indispensable tool in improving accountability of people responsible for delivering public services. Started in 2004 with nothing more than a shoe-string budget and plenty of enthusiasm, Lokvani has become a tool that enables people to track the management of civic amenities at the district level. The model is built around an information portal that is monitored by the district administrator's office on a daily basis. Designed to meet the unique needs of the local population, it provides a no-frills mechanism for people to interface with public officers in a simple and transparent manner via the telephone, which has a higher penetration than PCs or the internet. Moreover, the system remains sustainable by enabling local individuals and small businesses to set up licensed kiosks that provide access to the portal.

Given our current economic environment, it's no secret that unemployment is on the rise. While others face slowdowns, state agencies tasked with processing unemployment cases are flooded with claims. In Colorado, residents must go through a claims process for unemployment to verify continued benefits eligibility every 2 weeks. Colorado Department of Labor and Employment (CDLE) cannot afford the number of agents needed to handle the amount of calls received and knew it needed a self-service solution. CDLE determined that it was time to move on and in the end chose speech. Learn about how CDLE made its decision, the design and deployment methods used, the positive feedback received, and lessons learned.

In today's economy, everyone is concerned with cutting costs, but the best organizations practice customer-centric cost containment. They are using speech analytics to analyze customer interactions to cut costs with minimal impact on customer experience by finding the key drivers of customer satisfaction, so that they can focus investments on the activities that customers value most. In this session, attendees will learn how Pitney Bowes created an organizational model and process around the use of speech analytics in the contact center and leveraged the voice of the customer to cut costs.

Most companies are aware that their customers face daily challenges using the goods and services they offer. However a majority of them are not acutely aware of what those challenges are and how often they are impacting customers. Speech analytics is a valuable tool that can be used to identify and quantify the challenges faced by customers who in turn will generate actionable feedback to improve business processes and automation systems. In this session, attendees will learn how Cricket Communications proactively used speech analytics to identify and quantify areas of customer confusion to improve business processes, streamline automation, and improve customer satisfaction.