224s.09.lec13

224s.09.lec13 - CS224S/LING281 SpeechRecognition,Synthesis,...

Info iconThis preview shows pages 1–13. Sign up to view the full content.

View Full Document Right Arrow Icon
CS 224S/LING 281  Speech Recognition, Synthesis,  and Dialogue Dan Jurafsky Lecture 13: Dialogue: Information  State Systems and Dialogue Act  Interpretation
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Outline Natural Language Understanding Natural Language Generation Information-State Models Dialogue-Act Detection Dialogue-Act Generation
Background image of page 2
Summary The Linguistics of Conversation Basic Conversational Agents ASR NLU Generation Dialogue Manager Dialogue Manager Design Finite State Frame-based Initiative: User, System, Mixed VoiceXML Advanced issues in NLU and Generation Information-State Dialogue-Act Detection Dialogue-Act Generation Evaluation Utility-based conversational agents  MDP, POMDP
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Natural Language Understanding
Background image of page 4
Semantics for a sentence LIST       FLIGHTS   ORIGIN Show me   flights      from Boston  DESTINATION         DEPARTDATE to San Francisco  on   Tuesday DEPARTTIME morning
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
HMMs for semantics Idea: use an HMM for semantics, just as  we did for ASR (and part-of-speech  tagging, etc) Hidden units: Semantic slot names Origin Destination Departure time Observations: Word sequences
Background image of page 6
HMM model of semantics -  Pieraccini et al (1991)
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Semantic HMM Goal of HMM model:  to compute labeling of semantic roles C =  c 1 ,c 2 ,…,c n  (C for ‘cases’ or ‘concepts’)  that is most probable given words W argma x C P ( C | W ) = argma x C P ( W | C ) P ( C ) P ( W ) = argm ax C P ( W | C ) P ( C ) = argma x C P ( w i | w i - 1 ... w 1 , C ) P ( w 1 | C ) i = 2 N P ( c i | c i - 1 ... c 1 ) i = 2 M
Background image of page 8
Semantic HMM  From previous slide: Assume simplification: Final form: = argma x C P ( w i | w i - 1 ... w i - N + 1 , c i ) i = 2 N P ( c i | c i - 1 ... c i - M + 1 ) i = 2 M = argma x C P ( w i | w i - 1 ... w 1 , C ) P ( w 1 | C ) i = 2 N P ( c i | c i - 1 ... c 1 ) i = 2 M P ( w i | w i - 1 ... w 1 , C ) = P ( w i | w i - 1 ,.. ., w i - N + 1 , c i ) P ( c i | c i - 1 ... c 1 , C ) = P ( c i | c i - 1 ,.. ., c i - M + 1 )
Background image of page 9

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Semi-HMMs Each hidden state Can generate multiple observations By contrast, a traditional HMM One observation per hidden state Need to loop to have multiple observations  with the same state label
Background image of page 10
Another way to do NLU:  Semantic Grammars  CFG in which the LHS of rules is a semantic  category: LIST -> show me | I want | can I see|… DEPARTTIME -> (after|around|before) HOUR                        | morning | afternoon | evening HOUR -> one|two|three…|twelve (am|pm) FLIGHTS -> (a) flight|flights ORIGIN -> from CITY DESTINATION -> to CITY CITY -> Boston | San Francisco | Denver | Washington
Background image of page 11

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
An example of a frame Show me morning flights from Boston to SF on 
Background image of page 12
Image of page 13
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 77

224s.09.lec13 - CS224S/LING281 SpeechRecognition,Synthesis,...

This preview shows document pages 1 - 13. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online