Capstone: Spoken Dialog Systems

CSc 59866 - CAPSTONE I *Prof. Esther Levin* *Fall 2005* Spoken Dialog Systems and Voice XML
Time:	MW 12:15 am – 1:45 pm	Place:	NAC 7-312
Professor:	Prof. Esther Levin	Office Hours:	Mon 2:00-4:00
Email:	esther@cs.ccny.cuny.edu	Phone:	212 650-5626

Description:

Spoken dialogue systems enable users to interact with computer systems via natural and intelligent dialogues, as they would with human agents. Development of such systems requires a wide range of speech and language technologies, including automatic speech recognition (ASR), to convert audio signals of human speech into text strings, natural language and dialogue processing (NLP), to determine the meanings and intentions of the recognized utterances and to generate a cooperative response to them, and text-to-speech synthesis (TTS), to convert the system utterance into actual speech output.

VoiceXML is the HTML of the voice web, the open standard markup language for voice applications. VoiceXML harnesses the massive web infrastructure developed for HTML to make it easy to create and deploy voice applications. Like HTML, VoiceXML has opened up huge business opportunities: the Economist even says that "VoiceXML could yet rescue telecoms carriers from their folly in stringing so much optical fibre around the world."

While HTML assumes a graphical web browser with display, keyboard, and mouse, VoiceXML assumes a voice browser with audio output, audio input, and keypad input. Audio input is handled by the voice browser's speech recognizer. Audio output consists both of recordings and speech synthesized by the voice browser's text-to-speech system.

VoiceXML takes advantage of several trends:

The growth of the World-Wide Web and of its capabilities.
Improvements in computer-based speech recognition and text-to-speech synthesis.
The spread of the WWW beyond the desktop computer

Project Scope:

We will be designing, developing, testing and deploying spoken dialog system for a variety of applications. In particular, we will be designing an Automatic Reader Advisor for New York Public Library - a system that can automatically provide information about books and complete book orders through natural spoken dialog.

Upon the successful completion of this project a student should be able to:

understand the main functional components of a typical spoken language processing system;
have a detailed knowledge of the basic elements of spoken language technology, such as grammatical formalisms, speech recognition, and speech understanding;
have practical experience of speech recognition technologies and of spoken dialogue system development using VoiceXML;
appreciate current research issues in spoken language technology and be aware of its commercial applications

Logistics:

In this project-based course, students are grouped into teams to work on projects involved with design, implementation and testing of spoken dialog systems. The capstone course will last two semesters. In the first semester, we will study key technologies involved in this multi-disciplinary field. The second semester will focus on implementation of exciting real-world dialog systems using the Voice XML platform.

The course material will be entirely self-contained

Requirements and Grading:

Fall 2005: There will be 5-8 assignments totaling 50%-80% of the final grade. Some of the assignments will be research to be presented in class. The rest of the points are for 1 or 2 exams. Attendance is mandatory.

Text:

McTear, Michael, Spoken Dialogue Technology - Towards the Conversational User Interface. Springer Verlag, 2004

Additional Book Resources:

D. Jurafsky and J.H. Martin, SPEECH and LANGUAGE PROCESSING: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition, Prentice-Hall, ISBN: 0-13-095069-6, 2000.

R. Duda, P. Hart, D. Stork, "Pattern Classification", second edition, 2000.

L.R. Rabiner and B.W. Juang, Fundamentals of Speech Recognition, Prentice-Hall, ISBN: 0-13-015157-2, 1993.

L.R. Rabiner and R.W. Schafer, Digital Processing of Speech Signals, Prentice-Hall, ISBN: 0-13-213603-1, 1978.

X. Huang, A. Acero, and H.W. Hon, Spoken Language Processing - A Guide to Theory, Algorithm, and System Development, Prentice Hall, ISBN: 0-13-022616-5, 2001.

Announcements:

· Aug 29-th: Assignment 1 is posted. Due Sep. 12th in class.

· Aug 29-th: No lab meeting on Wed Sep. 7^th

· Sep 12-th: Our next meeting (wed Sep 14-th) will be held in NAC 5/126

· Sep16-th. Assignment 2 is posted. Due Oct 3-rd.

· Sep 22-nd. Grades for Assignment 1 are posted

· Sep 24-th. Please complete the Leeds Online Tutorial on HMMs (except Forward and Forward/Backward Algorithm) before the lecture on Sep 28-th.

· Sep 25-th. : Assignment3 is posted. Due Oct 12-th.

· Sep 29-th: No lab meeting on Mon Oct 3^rd.

· Oct 4-th: Assugnment 4 is posted Due Oct 24-th.

· Oct 10-th. Grades for Assignment 2 are posted.

Syllabus (evolving and subject to change):

Weeks	Topic
1	Introduction to Dialog Systems
2-7	Pattern Recognition: Introduction to Pattern recognition, Bayesian Classifiers, Hidden Markov Models, Tutorial on HMMs
8-14	Spoken dialog systems, Speech recognition, Grammars, Dialog Managements, Spoken dialog Engineering.

Web Resources:

Natural Language Processing course from University of Ulster

VoiceXML

VoiceXML development platform: Bevocal Café

Other VoiceXML developer resources

W3C documentation

VoiceXML Forum

W3C Dialog Requirements for Voice Markup Languages (http://www.w3.org/TR/voice-dialog-reqs/)

Developer.com (Voice) (http://www.developer.com/voice/)

The XML Cover Pages VoiceXML Forum (Voice Extensible Markup Language Forum)

Voice Services: What sorts of voice applications are best suited for VoiceXML? Here are a few ideas. (http://www.voicexml.org/tutorials/intro6.html)

Sites with sample applications and demos:

Nuance Communications: http://www.nuance.com
Apple: http://www.apple.com/macos/speech/
Scansoft: http://www.scansoft.com/

Testing VoiceXML applications