Designed for attendees new to the speech technology arena, this tutorial provides an overview of today’s key speech technologies. What are the major differences between dictation and conversational recognition engines? Should you use statistical language models (SLMs) or speech grammars, and do they need semantic tags? Should you use concatenative or parameter based speech synthesis engines? Do speaker identification and verification really work? Is there a need for touchtone recognition in interactive voice response systems? Who drives the speech dialogue? The user, the computer, or both? Where can natural language processing technologies such as natural language recognition, machine translation, response generation, and summarization be used today? Can speech technologies be embedded in mobile devices, or do they require a backend server? Should you use POTS, VoIP, or some other communication strategy? Untangle the voice standards alphabet: VoiceXML, SSML, SRGS, CCXML, PLS, and SCXML.

Jump-start your knowledge in the field of voice user interface design! This session is designed to quickly get those new to VUI design up to speed so they can make the most of the Principles of VUI Design track at the conference. This tutorial will illustrate why VUI design is the make or break factor for speech applications and how to make smart design decisions from Day 1. Learn how to encourage customers to accept and use speech automation by focusing on the perceptions and reactions of end users throughout the design process. This tutorial will cover the basics in VUI design: technology current and future state (including multimodality), speech project methodology, design principles and caller's rules for efficient, no-nonsense call flows; and evaluation techniques to learn what works and what real customers think.

Back by popular demand! This course (first offered in 2008) will get you going with grXML grammars. The first half will start with a quick overview of grXML structure so we can analyze sample grammars, then look at guiding principles and best practices. Plus, we’ll examine some cases when applying best practices really isn't best (there's always a loophole!). Once you have basic grammars down, the second half will explore testing, tuning, and how you can add cool, advanced stuff to further constrain your grammar and optimize recognition. Whether you’re a VUI designer, developer, manager, analyst, QA specialist, or anybody else, if things like lexicons, rulerefs, OOG, and FA-out are a bit fuzzy to you, or you just want to better understand how to use grammars to boost performance of your speech applications, this course is for you.

Natural language (NL) processing is supporting increasingly sophisticated speech dialogues. This session will introduce NL processing and its role in current and future speech applications to understand what NL is, techniques for processing NL, and where NL processing is heading. We will describe how to use statistical language models (SLMs) and speech grammars with semantic tags for processing NL. Attendees will develop tagged grammars illustrating the techniques discussed in the tutorial. We will also discuss the roles of emerging standards and research in future applications and how standards such as Extensible MultiModal Annotation (EMMA) will contribute to more sophisticated applications. This session will conclude with a brief look at some recent topics in NL research.

Businesses employing speech technology depend on a successful match between their implementations and the natural human capacity to understand and use speech. When the two align, the results can delight. When they don’t, the results can be increased failures and reduced caller satisfaction. Linguistics, the scientific study of natural language, can provide helpful insights into how to create, evaluate, and improve voice user interfaces (VUIs) so that users’ innate abilities and expectations align with what is experienced. This tutorial will introduce participants to key linguistic principles and applied methods that have high impact on the caller experience and business outcomes. Technologists, business owners, and VUI designers will all gain insights into pain points that can be better understood and resolved through tricks of the linguistic trade. Prior to the tutorial, participants will receive a set of activities to be completed in advance. Completion of activities is not mandatory but highly encouraged.

This course will teach participants a rigorous, data-driven speech application tuning methodology that will enable them to build robust speech applications that effectively deal with how real users actually behave, not how we would like them to behave. Topics include utterance and dialogue-level performance metrics, managing out-of-grammar utterances, techniques to effectively identify and address performance problems, dealing with multitoken utterances, tuning phonetic dictionaries, computing enhanced confidence scores, setting confidence thresholds, and running dialogue simulations. The presentation will be illustrated by numerous examples and interactive demonstrations using field data from real-life applications.

Ensuring the success or failure of a speech application requires determining the relevant success criteria, measuring the performance of the application based on those metrics, and taking the appropriate actions at each stage of development to ensure the criteria are met. This tutorial will discuss how to determine success criteria, based on the business goals of the project; benchmarking the application metrics to ensure the success criteria are appropriate and achievable; evaluating the performance of competing application prototypes; and ensuring continued application success after project completion through continuous improvement.

A computer telephony integration (CTI) application offers businesses the ability to increase profits by not only lowering expenses, but also, and perhaps more importantly, through brand differentiation and increased customer loyalty. Implementing CTI has numerous benefits to the enterprise and the caller. It is able to dramatically improve customer satisfaction while simultaneously reducing call center costs and improve security by eliminating a potential breach point. The presentation will define CTI, outline its benefits and impact on the caller experience, and quantify the return on investment.

The success of a speech solution often depends on whether the right requirements are gathered and captured in a useful way. An application can be “built to spec” and still not thrive if requirements and success metrics are ill-conceived. This hands-on workshop will present specific techniques for capturing the user-based requirements more likely to lead to successful project outcomes. Benefits of this approach include the ability to do the following: differentiate various types of users who will interact with the speech solution; create contextual stories around the high-level goals for each type of user; break down high-level goals into manageable requirements; and specify acceptance criteria so stakeholders can agree when requirements are met. This workshop builds upon Mike Cohn’s (2004) work on User Roles and User Stories, which will be introduced and applied to speech projects through group exercises.

The proliferation of the internet and digital applications has had many ramifications — a significant one being a spurt in faceless electronic transactions and fraudulent activities such as identity theft. The majority of identity theft, which is on the rise, happens in the non-internet world. Abuse also happens in the digital world, with significant attempts being made to prevent fraudulent online activity. Voice biometrics assists in the perpetration of false identities and makes the hurdle very difficult to overcome. The session will include a live demonstration for speaker verification in which randomly selected speakers can enroll their voices and verify against their voiceprints. Also covered will be the basics of digital signal processing and speaker identification; how noise filters work and how to account for various channels; the effectiveness of voice verification methods; drawing a balance between security and convenience; and the basics of voiceprint representation.