LING 575 - Spoken Dialog Systems
Spring 2016
Specialized Topics

NOTE: You may work in pairs or individually on the presentation.

• A core/primary reading for each topic will be identified on the associated topic page. Everyone is expected to have read that paper prior to class. At least 2-3 other major papers are identified, and other supporting papers may also be posted. Presenters should read those papers, and are also encouraged to find other related materials. The ACL anthology is an excellent resource, as is the SIGDial web site.

To-Dos

• Participants
• Read at minimum the primary reading in advance of class. Reading the secondary articles is, of course, encouraged.
• For each week, please choose one of the primary papers as the focus of a critical reading assignment.
For the critical reading assignment, you should write a (roughy) one page response to the paper and identify questions for class discussion. You might find it useful to read Mitzenmacher and Ramsey's guidelines on reading research papers (thanks to Jan Wiebe and Rebecca Hwa for the pointer).
• Your write-up should include:
  • A brief summary of the main point of the paper.
  • Notes about the interest of the paper from an NLP perspective
  • At least one observation about: its relation to related work, strengths or weaknesses, ways it could be improved, ways it could be extended, or how it could be used to improve another application.
• You shoud post to the corresponding GoPost conversation:
  • One or more questions or issues for class discussion.
• The assignment should be submitted by 11:45p.m. Wednesdays, but submitting earlier will allow me and the presenters to review the comments prior to class.
• Presenters
  • Presenters should plan to present for 10-15 minutes, with an additional 10 minutes for discussion.
  • The members of the class will also have read the paper, so you do not need to detail all the content in the paper(s). You can focus on the primary paper to lay out main concepts, and use the other papers to provide contrasts or supplementary examples. Instead, you should present a synthesis and analysis of the topic, abstracting away from the gory details:
  • What problems are being addressed? Why are they interesting/challenging?
  • What are the goals and main claims of the paper(s)?
  • What sorts of data/approaches/evaluation are employed?
  • What are the main things you've learned about the topic?
  • Provide a *constructive* critique.
  Your goal is also to stimulate and lead discussion of the topic.
  • How feasible are these ideas to implement in a dialog system?
  • What would be the effect of implementing these sorts of techniques in dialog systems?
  • Are the approaches more motivated by modeling human conversational behavior or engineering considerations?
  • Presentation material (slides) may be in PowerPoint, PDF, Google Docs, etc. It would be appreciated if the slides were provided prior to the start of class.
  • Online presenters should make sure to have a microphone available. It is recommended that online presenters enter the meeting room a few minutes before class to allow for checking of audio levels.

List of possible specialized topics:

>

Date	Presenter(s)	Topic
April 28
April 28	John M.	Recognition and understanding of prosody
April 28	Jeff	Turn-taking and Backchannels
April 28	Eslam	Incremental Processing of Dialog
April 28	Joanna	Multi-party dialog
May 5
May 5	Nate	Recognizing user affect in human-computer dialog
May 5	Katherine	Modelling affect in dialog
May 5	Micaela	Sentiment and subjectivity in dialog
May 5	Gilsinia	Deep Learning for Dialog Systems
May 12
May 12	Jason	Dialog act taxonomies
May 12	Ernie	Dialog act recognition
May 12	Laurie & George	Miscommunication and Repairs
May 12	Thor	Grounding
May 19
May 19	Charlie	Information structure and prosody in dialog
May 19	Vinay, Ethan	Reference; anaphora & generation in dialog
May 19	Maria	Mixed Initiative
May 26
May 26	Lauren	Persona and Personification in dialog
May 26	Elizabeth	Dialog with Different User Populations
May 26	Spencer	Dialog genres: information-seeking, tutoring, conversational
May 26	Veljko	Dialog in other NLP applications: search, translation, question-answering