1 / 6

Human Act of Feedback ”When should I backchannel/barge-in?”

Human Act of Feedback ”When should I backchannel/barge-in?” . Ling575 Spoken Dialog Systems June 5 , 2013 Sanae Sato. Some Experiment…. Natural human-human conversation Three people / 60 min / roughly 970 sentences. Context: dinner table conversation

kaycee
Download Presentation

Human Act of Feedback ”When should I backchannel/barge-in?”

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Human Act of Feedback”When should I backchannel/barge-in?” Ling575 Spoken Dialog Systems June 5, 2013 Sanae Sato

  2. Some Experiment… • Natural human-human conversation • Three people / 60 min / roughly 970 sentences. • Context: dinner table conversation • Back-channeling/Barge-in are used quite effectively among human-to-human conversation. (= 3-5 % of conversation)

  3. Human Act of Feedback • From Observation… • Both backchannel and barge-ins plays important role for grounding purpose. Especially, Barge-in is extremely powerful for listener to keep track of topic/context. • Assumption1: Generating backchannelsis effective in term of grounding. • Assumption2: Similarly, generating effective barge-ins can lead to more beneficial interactions. • “Did you mean say Indian or Italian?”  Increase confd. Score • Another scenario: Reaction to barge-in  Yield of keep the turn? • Silence?: not considered for this project. • Question: When should I (or system) backchannel/barge-in?

  4. Main Topics to be Covered • Framework Model: • Dethlefset al. 2012“Optimising Incremental Dialogue Decisions Using Information Density for Interactive Systems” • Domain: Restaurant Recommendation Application • Information Presentation • Hierarchical Learning • Incremental Actions • Semi-Markov Decision Process (SMDP) • Slot-ordering & Confidence Score • …Good paper, but it lacks of human data!

  5. Main Topics to be Covered • Information Density (Jaeger 2010) • Dialogue Management • Turn-Taking • Data: • Dinner table conversation • CALLHOME (telephone speech corpus)

  6. References • Nina Dethlefs, Helen Hastie 2012 Optimising Incremental Dialogue Decisions Using Information Density for Interactive Systems. @ ACL 2012 • T. Florian Jaeger 2010 Redundancy and reduction: Speakers manage sysntactic information density. (Cognitive Psychology 61:23-62) • From Class Reading: • David Schlangen & Gabriel Skantze A General, Abstract Model of Incremental Dialogue Processing @ European ACL 2009 • Gabriel Skantze & David Schlangen:IncrementalDialgueProcessinginaMicro-Domain @ European ACL 2009 • David DeVault, Kenji Sage, and David Traum. 2009. Can I finish? Learning when to respond to incremental interpretations results in interactive dialogue. In Proc. SIGdial 2009 Conference, pages 11-20 • Ethan O. Selfridge, et al. 2011. Stability and accuracy in Incremental Speech Recognition

More Related