Saturday, December 29, 2012

Conversation Handouts: Dialogue and Conversational Agents

Jurafsky & Martin Speech and Language Processing Ch 19


Outline
  • why dialog is different 
  • representing and interpreting dialog acts 
  • dialogue structure'and coherence

Next time:
  • dialogue system architecture
  • dialogue management and, other modules 
  • case study of Why-Atlas'tutoring system and 
  • NLP in industry .

Dialogue and Conversational Agents
  • "Conversation or dialogue is the most  fundamental and-specially privileged arena of language."
  • What makes dialogue different?
    • turn-taking
    • grounding 
    • Imptlicature
  • Speech act representation and interpretation.
  • Approaches to Coherence and structure.
Example Conversational Agents
  • Phone system for flight reservations
  • Why2 Physics Tutor
  • NJFUN ---> is a real-tim spoken dialogue system that provided users with information about things to do in New Jersey. NJFUN is built using an internal platformal for spoken dialogue systems. NJFUN uses a speech reccognizer with stochastic language models trained from example user utterances, and a TTS system based on concatenative diphthone synthesis. Its database is populated from the NJ-online webpage to contain information about activities. NJFUN indexes this database using three attributes: activity type, location and time of day. 
Turns and Utterances
  • Dialogue is characterized by turn-taking.
  • Speakers know how to take turns (who should talk next, and when they should talk)
    • little overlap (around 5% in English - although depends on the domain)
    • not much silence between turns either.       
Turns:
  • Conversational Analysis provides a socio-linguitic approach to turn-taking (e.g. Sacks et al)
  • Transition-relevance places are where the structure of the language allows speaker shifts to occur.
  • Turn-Taking Rule (simplified): At each transition-relevance place of each turn:
    • If during this turn current speaker has selected A as the next speaker, then A must speak next.
    • If current speaker does not select the next speaker, any other speaker may take the next turn.
    • If no one else takes the next turn, the current speaker may take the next turn.
Utterances:
  • Transition-relevance places are typically at utterance boundaries. 
  • Spoken utterances are typically shorter, contain more pronouns, have repairs compared to written sentences.
  • Many theories take the utterance as the primitive unit, but utterances are difficult to segment.
    • a single utterances may occur across several turns.
      • A: We've go you on USAir flight 99.
      • B: Yep
      • A: leaving on December 1.
  • Multiple utterances may occur in a single turn. 
    • We've got you on USAir flight 99 leaving on December 1. Do you need a rental car?
  • Linguistic boundary clues include words, ngrams, prosody.      
Conversation Analysis (Cont.)
  • GREETING            GREETING
  • QUESTION          ANSWER
  • COMPLIMENT   DOWNPLAYER     
  • REQUEST            GRANT
Significant silence (follows first part of an adjacency pair).
  • A; Is there something bothering you or not?  (1.0)
  • A: Yes or no?   (1.5)
  • A: Eh?
  • B: No.
Implications for spoken dialogue systems