This session offers a candid debate on the state of speech patents, both as assets and tools for the commercial market- place, and changes in the process for obtaining patents—especially in view of the America Invents Act. How have these provisions affected filings for speech patents, re-examination proceedings, and litigation activity? What pitfalls exist for the unwary? The discussion also includes a dissection of a recently issued speech patent so attendees can see how patents are actually made.

Alphanumeric speech recognition is one of the more challenging problems we face: There are many possible ways to implement it. When recognition goes well, callers are more satisfied; when recognition is not good, callers can get frustrated. This session covers how to design for improved alphanumeric recognition. It also includes various tips and tricks—constraining grammars versus using an n-best list, using web services to validate input, leveraging prefiller to improve initial phoneme recognition, and how to compare recognition methods.

Sentiment analysis can provide insights to businesses, academics, nonprofits, political parties, and of course government intelligence agencies about the emotions and sentiments of speakers and text writers. What is the role of speech technology in sentiment analysis? Can we get more than just “speech to text”? What will be the role of the new W3C EmotionML language?

This session discusses current and emerging standards in speech and related areas. Questions for discussion include the following: What standards are relevant to different types of speech applications, for example, traditional IVRs, desktop browsers, and mobile apps? What standards are relevant to developers and what standards are relevant for platform implementers? What platforms implement the various standards? How can we learn more about the different standards? How can we participate in standards development?

In an increasingly competitive environment, customer service can be the difference between winning and losing customers and also turning neutral customers into some of your biggest advocates. To do this, organizations need not only to meet customer expectations, but exceed them. Adapted from Bliss’s best-selling book, I Love You More Than My Dog; Five Decisions That Drive Extreme Customer Loyalty in Good Times and Bad, this keynote focuses on what’s most important for delivering meaningful customer experiences that drive profitability and growth. It’s not about a point in time; it’s about deliberate decisions and actions that lead to customer rave and earn customer advocates.

Smartphones and other mobile devices provide continuous access to web resources and to our contacts. Speech and natural language technology make it easier to use these devices. The combination makes possible a constant and tighter connection between computers and individuals—a significant technology and business trend. There are major implications for businesses, both for consumer-facing and for internal enterprise operations. This talk examines the language technology and supporting infrastructure developments required to make the most of this trend.

The choice between cloud-based and on-premise speech recognition has faced organizations for many years, but the nature of customer interaction is changing as customers expect intelligent and consistent phone, mobile, and web interactions. Join us for a discussion of the features, benefits, and limitations, as well as real-world applications of cloud, on-premise, and blended speech recognition to help you reinvent your customer experience.

Recent trends in customer behavior reveal that customer contacts can be segmented according to how often a customer contacts a company. The combination of frequency and regularity of contact patterns allows us to create a model that will allow organizations to develop an effective customer service strategy for both self- and assisted-service across various channels.

By the end of 2013, approximately half of all businesses are expected to use speech analytics as a part of their customer care efforts. Research findings demonstrate that real-time speech analytics can empower agents with crucial intelligence on customer behavior at the moment of truth—during the call. This session provides an overview of how real-time processing of speech data enables organizations to outperform their competitors in key measures such as customer satisfaction and average handle time.

Join us for a survey of the voice biometrics ecosystem, including the role of Big Data and behavioral analytics, as it is being applied in call centers and mobile customer care systems today. Learn where we stand with voice biometrics, both in terms of the technology and market maturity curves, and hear an analysis of the overall potential for voice biometrics to support better conversational commerce.

A menu can make or break your IVR. Join the AVIxD Design Guidelines Working Group for a brief overview of guidelines on menu structure, wording, and recording, followed by the chance to get your hands dirty applying the guidelines in tricky situations. Bring your own menu challenges to the session for the opportunity to work together on solid solutions.

Good VUI design needs to anticipate all the possible things that can go wrong in the interaction between end users and speech applications. Designers must create elegant error recovery strategies to minimize the impact of errors and allow the interaction to continue. This session examines the effectiveness of several common techniques such as confirmations and escalating reprompt strategies and other approaches. Real caller data, including audio and video, gives participants a fresh, hands-on look into this important design topic.

Prompt writing is the absolute heart of voice user interface design. A select few designers are born with the ability to write a clear, conversational prompt; the rest of us could benefit from a set of well-defined prompt writing guidelines. The first part of this session showcases critical prompt writing tips from the AVIxD Design Guidelines Working Group. During the second part of the session, we break into small, interactive groups to tackle real- world prompt issues.

An unwritten assumption in VUI design is that minimizing the time callers spend in the IVR will result in improved caller satisfaction. In this session, we question this assumption and present evidence that suggests that the actual time callers spend resolving an issue is less important than their perception of whether they are making progress. Bolstered by studies from behavior sciences, we argue that designers should focus on minimizing the caller's perception of waiting and maximizing their sense of progress.

Prompts that request information from callers are vital to the success of most speech applications. This session presents strategies from the AVIxD Design Guidelines Working Group for collecting all- digit data elements (phone numbers, credit card numbers, dates, and amounts), as well as more complex data elements such as addresses, names, email addresses, and alphanumeric information such as Canadian postal codes. Bring your own data-collection problems and work together to find solutions.

Nearly three-quarters (74%) of today’s customers use at least three channels when interacting with a company for customer service, according to Ovum. Sadly, most companies are still serving stand-alone, single channel experiences that frustrate customers by losing the thread of the interaction and forcing customers to repeat themselves as they move from one channel to another. This session provides a strategy for using Big Data to predict customer intent and capturing the interaction context to enable a unified, channel-agnostic digital experience.

The connected car has revolutionized the driving experience with advances in communication, entertainment, and navigation. However, there are concerns about driver distraction, emphasizing the need for natural interfaces involving speech and gesture. What are the key ingredients of a safe user interface for secondary driving tasks? How close are we to having robust virtual assistants that are safe to use while driving? In this session, we also explore hybrid solutions where intelligence resides in the vehicle and in the cloud as well as how the business models need to be structured.

Driving presents a complex environment for a speech system. Grammars and statistical models must reflect partial commands and location- or trip-specific queries. Pronunciations must account for the unusual such as De Guigne Drive. Dialogue must also be innovative, at the macro level and the micro level of turn-taking and barge-in. The system needs to gracefully initiate a conversation as well. This presentation demonstrates and explains this next-generation approach to the speech experience while navigating.

Contact centers and app developers are turning to voice biometrics to improve security while enabling customers to interact with companies in a hands-free, convenient manner. Voice-based authentication of callers is key to secure, trusted customer care interactions. With the right testing plan, enterprises can ensure that their applications for voiceprint enrollment as well as ongoing fraud detection are fast and friction-free.

Companies are striving to ensure that a good customer experience and inexpensive customer service are not mutually exclusive. The key to this approach is to identify functions for which customers are happy to use automation, but discovering and validating these opportunities requires a thoughtful process. This talk presents a proven customer experience assessment approach that puts the customer first by identifying customer needs through observation in the call center, uncovering business goals through stakeholder interviews, and ultimately balancing customer needs with business goals.

See two 20-minute presentations about some of the latest technological developments in speech technology from companies at the conference. The specifics of this presentation will be available on www.SpeechTEK.com. All session updates will be included in the final program distributed at the SpeechTEK 2013 conference. Also, to see what conference organizers, speakers, and attendees are saying about this event, follow #SpeechTEK on Twitter. To join the SpeechTEK 2013 group discussion, include #SpeechTEK in your Tweets.

Not only are PCs, TVs, and gaming consoles connected to the internet, soon literally hundreds of devices in the home will be connected—thermostats, light switches, security systems, irrigation systems, refrigerators, and washing machines. Will voice be part of the user interface to these devices and will voice recognition be embedded or cloud-based? Will control continue to reside with individual devices, or will it be integrated into a universal remote control or mobile device? What APIs are needed for controlling the many devices in the home? Find out in this session.

Learn about the challenges associated with keyword spotting within large speech databases. Target markets are described in this session along with the specific technology requirements of each. Current keyword spotting mechanisms are presented and compared as well, with a focus on the efficient implementation of phonetic searching.

We're introducing Rivr, an open-source Java dialogue engine for building high-quality, high-performance voice applications. At the core of the engine is an intuitive, easy-to-use yet powerful programming model. It has been used to build some of the most sophisticated IVR applications on the market. Its design is centered on many essential characteristics: the ability to design truly reusable dialogue strategies, testability, and real support for devel- oping enterprise applications.

How can companies have their customers connect to speech systems and to customer service agents for assistance directly from within a mobile app or a PC browser session? With no heavy downloads or application installations, WebRTC provides voice, video, and cobrowsing so that an automated or live agent can help immediately from within the context of the customer’s mobile or browser activity.

For unknown domains or where training on contextual data is impractical, generic statistical language models (SLMs) provide accurate speech recognition but are not easily modified to suit specific applications. We describe a class of automatic speech recognition solutions, based on generic SLMs, where the developer has an option to improve accuracy by providing configuration hints at runtime in the form of grammars or vocabulary lists. We cite performance examples and case studies from live deployments.

See two 20-minute presentations about some of the latest technological developments in speech technology from companies at the conference. The specifics of this presentation will be available on www.SpeechTEK.com. All session updates will be included in the final program distributed at the SpeechTEK 2013 conference. Also, to see what conference organizers, speakers, and attendees are saying about this event, follow #SpeechTEK on Twitter. To join the SpeechTEK 2013 group discussion, include #SpeechTEK in your Tweets.

Join your peers on Monday after the sessions for the opening of the Customer Solutions Expo - featuring the CRM Evolution, SpeechTEK, and Customer Service Experience Showcases. Visit with conference sponsors, exhibitors, speakers, and other attendees while enjoying light hors d'oeuvres and drinks.