Customer Reviews

How to Build a Speech Recognition Application

5 star:		(5)
4 star:		(1)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

Large vocabulary recognition applications are hot. Callers can speak directly to computers to make reservations, trade stocks, track packages, and handle many other complex transactions. Convenient applications that allow you to use the telephone to interact with computerized tranaction systems will soon be ubiquitous. Designers with the experience and cleverness to engineer good applications are in demand.

This book provides an excellent guide for both the novice and experienced application designer. Balentine and Morgan are both well known and well respected in the User Interface and Speech Processing technical communities.

The book covers a wide variety of topics useful in creating speech recognition applications: speech output, speech recognition, natural language understanding, vocabulary suggestions, testing, etc. All of the topics are covered from an intensely practical approach. The authors provide lots of guidelines for application dialog definition, and distinguish required from recommended practices. They make an effort to point out good design practices. A careful reader of this book can not help but become a better dialog designer and application developer.

But this book is not just a cookbook, or a list of guidelines. Balentine and Morgan make sure the reader understands why a recommendation is made. In difficult or controversial areas, they discuss important issues that affect the application dialog. For example, in the chapter on barge-in they avoid the simple minded approach used by some technology vendors and discuss when barge-in should and should not be used, how to recover from error, and the legal issues posed.

I have spent 15 years working with ASR and Text-to-Speech technologies from the perspective of an application designer and system implementer- this is the first book that has provided a complete "education" on the "how to use these technology advances with the greatest chance of actual user acceptance and assistance in doing it right the first time out". This book is a primer for first time technology interest, and a must-have for those technical types who implement automation systems on an on-going basis.

Author, Bruce Balentine's goal with How to Build a Speech Recognition Application is to produce the Strunk and White of speech recognition. An electronic musician and composer, Balentine was a pioneer in the speech recognition field. The text's examination of the problems of navigation from human perception to machine recognition give comprehension to even the layman. The book is well organized and structured with the +,/,- system which allows a novice to follow. According to linguist, Dr. John White, the first chapter could stand alone as a treatise on the dynamics of the human speech interface with the machine.

This book is a definitive resource for any professional seeking to design a clean, usable and rock-solid telephony interface. In fact, the book takes you beyond speech recognition per se and helps you to understand the logic and heuristics of good telephone interface design. I highly recommend this text for anyone who is putting together a new system or has to re-do an existing one.

Developing speech applications is not easy to master. Even with VoiceXML becoming more widely adopted, there are a lot of intricacies that that a developer must understand. This book will provide you with a solid foundation to become an effective speech application developer.

The book did very well in presenting the limitations of the current speech recognition technology (dialog design, large vocabularies, promtp design, etc.) and made suggestions on how to overcome such problems in specific situations.

When this book came out a couple of years ago it was the first and only book on designing speech recognition systems. It was very valuable then, but now more books are available that cover the same information and more, for a lot less money.