Spoken Dialogue Systems (LSA 353), Regular Session, 5-27 July


TTh 10-12




Julia Hirschberg

Office Hours: 



julia [at] cs.columbia.edu












We will discuss current issues in spoken dialogue systems, including the modeling of human-human behavior such as turn-taking, grounding, timing, lexical entrainment, feedback, and clarification strategies; automatic speech act identification, error detection and correction strategies; and the technologies supporting such systems and how we evaluate them. Examples will be drawn from commercial and research spoken dialogue systems. This course is suitable for all levels and will be conducted as lecture/discussion.

Course Areas: Computational Linguistics, Discourse


All are available on line, linked to this syllabus.

Academic Integrity:

Copying or paraphrasing someone's work (code included), or permitting your own work to be copied or paraphrased, even if only in part, is not allowed, and will result in an automatic grade of 0 for the entire assignment or exam in which the copying or paraphrasing was done. Your grade should reflect your own work. If you believe you are going to have trouble completing an assignment, please talk to the professor in advance of the due date.




TTS Systems, more TTS Systems, and still more





Readings and Assignments

Reports and HW


F Jul 6

Issues in Spoken Dialogue Systems




Tu Jul 10

When to Start and When to Stop

 J&M 22.1 (new version); Clark03; Beattie82



F Jul 13

Components of SDS: ASR, TTS

J&M 22.2 (new version)



Tu Jul 17

Building and Evaluating SDS

J&M 22.4,8 (new version); Walkeretal97



F Jul 20

Generating and Detecting Dialogue Acts

J&M 22.5 (new version);Jurafsky98, Rosset&Lamel04



T Jul 24

Error Detection and Repair Strategies

J&M 22.5 (new version); Hirschbergetal04, Goldberg03, Krahmer01



F Jul 27

SDS Future

 J&M 22.2 (new version); Brennan96; Roth05


Course Paper Assignment:

A term paper, 8-10 pages long, on the following topic:

Eliza is a classic text-based AI Dialogue System which often fools casual users into thinking they are conversing with a human being.  For this assignment:
  • Become acquainted with the text Eliza at this link.
  • Identify 5 or more strategies you can use to demonstrate that Eliza is not a human conversationalist (e.g. linguistic constructions she does not handle well, pragmatic behaviors she does not generate appropriately or recognize).  As evidence describe your inputs and her outputs
  • Design a speech-enabled version of Eliza:
  • Explain the difficulties you will have to overcome in recognition, generation, and dialogue management, compared to the text version, based upon the topics we have studied in class and additional observations you may have
  • Suggest specific ways of dealing with these problems based upon what you know of the state of the art in SDS or ideas you may have to improve upon it
  • Indicate the ways in which a spoken version of Eliza might be even better than the text version, in terms of how it might interact with the user and the features of spoken language you might be able to make use of in generation and recognition of user behavior