CSC 475592 Natural Language Processing

1
CSC 475/592Natural Language Processing

• Dr. Curry I. Guinn
• MW 330-445
• CI 2006

2
Today

• Chapter 18.2-18.3
• Chapter 20

3
Text Coherence

• Example
• (1) John hid Bills car keys.
• (2) He was drunk.
• (1) John hid Bills car keys.
• (2) He likes junk food.
• (1) George Bush supports big business.
• (2) Hes sure to veto House Bill 1711.
• Hearers try to find connections between
  utterances in a discourse.
• The possible connections between utterances can
  be specified as a set of coherence relations.

4
Coherence relations (Hobbs,1979)

• Result S0 causes S1
• John bought an Acura. His father went ballistic.
• Explanation S1 causes S0.
• John hid Bills car keys. He was drunk.
• Parallel S0 and S1 are parallel.
• John bought an Acura. Bill bought a BMW.
• Elaboration S1 is an elaboration of S0.
• John bought an Acura this weekend. He purchased
  it for 40 thousand dollars.

5
Discourse structure

• S1 John took a train to Bills car dealership.
• S2 He needed to buy a car.
• S3 The company he works for now isnt near any
  public transportation.
• S4He also wanted to talk to Bill about their
  softball leagues.

Explanation
6
Discourse structure

• S1 John took a train to Bills car dealership.
• S2 He needed to buy a car.
• S3 The company he works for now isnt near any
  public transportation.
• S4He also wanted to talk to Bill about their
  softball leagues.

Explanation
Parallel
7
Discourse structure

• S1 John took a train to Bills car dealership.
• S2 He needed to buy a car.
• S3 The company he works for now isnt near any
  public transportation.
• S4He also wanted to talk to Bill about their
  softball leagues.

Explanation

Explanation
Parallel
8
Discourse parsing
Explanation (e1)
S1 (e1)
Parallel (e2e4)
Explanation (e2)
S4 (e4)
S2(e2)
S3(e3)
9
Why compute discourse structure?

• Natural language understanding
• Summarization
• Information retrieval
• Natural language Generation
• Reference resolution

10
Two theories on discourse structure

• Mann and Thompsons Rhetorical structure theory
  (1988)
• Grosz and Sidners attention, intention and
  structure of discourse (1986)

11
Rhetorical structure theory (RST)

• Mann and Thompson (1988)
• One theory of discourse structure, based on
  identifying relations between parts of the text
• Defined 20 rhetorical relations
• Presentational relations intentional
• Subject matter relations informational
• Nucleus central segment of text
• Satellite more peripheral segment
• Relation definitions and more.

12
Some examples

• Explanation John went to the coffee shop. He was
  sleepy.
• Elaboration John likes coffee. He drinks it
  every day.
• Contrast John likes coffee. Mary hates it.

13
Discourse structure
John likes coffee
They argue a lot
contrast
cause
elaboration
Mary hates coffee.
He drinks it every day
14
A relation Evidence

• (a) George Bush supports big business.
• (b) Hes sure to veto House Bill 1711.
• Relation Name Evidence
• Constraints on Nucl H might not believe Nucl to
  a degree satisfactory to S.
• Constraints on Sat H believes Sat or will find
  it credible
• Constraints on NuclSat Hs comprehending Sat in
  Sat increases Hs belief of Nucl.
• Effect Hs belief of Nucl is increased.

15
A relation Volitional-Cause

• (a) George Bush supports big business.
• (b) Hes sure to veto House Bill 1711.
• Relation Name Volitional-Cause
• Constraints on Nucl presents a volitional action
  
• Constraints on Sat none.
• Constraints on NuclSat Sat presents a situation
  that could have caused the agent of the
  volitional action in Nucl to perform the action.
• Effect H recognizes the situation presented in
  Sat as a cause for the volitional action
  presented in Nucl.

16
Another example

• S (a) Come home by 500. (b) Then we can go to
  the hardware store before it closes. (c) That way
  we can finish the bookshelves tonight.
• (a)
• (a) (b)
  (c)

motivation
motivation
(b)
(c)
condition
condition
17
Problems with RST (Moore Pollack, 1992)

• How many rhetorical relations are there?
• How can we use RST in dialogues?
• How do we incorporate speaker intentions into
  RST?
• RST does not allow for multiple relations between
  parts of a discourse informational and
  intentional levels must coexist.

18
Grosz Sidner (1986)
19
Grosz and Sidner (1986)

• A leading theory of discourse structure
• Three components
• A linguistic structure
• An intentional structure
• An attentional state

20
Linguistic structure

• The structure of the sequence of utterances that
  comprises a discourse.
• Utterances form Discourse Segment (DS) and a
  discourse is made up of embedded DSs.
• What exactly is a DS?
• Any evidence that humans naturally recognize
  segment boundaries?
• Do humans agree on segment boundaries?
• How to find the boundaries automatically?

21
Intentional structure

• Speakers in a discourse may have many intentions
  public or private.
• Discourse purpose (DP) the intention that
  underlies engaging in a discourse.
• Discourse segment purpose (DSP) the purpose a
  DS. How this segment contributes to achieving the
  overall DP?
• Two relations between DSPs
• Dominance if DSP1 contributes to DSP2, we say
  DSP2 dominates DSP1.
• Satisfaction-precedence DSP1 must be satisfied
  before DSP2.

22
Attentional State

• The attentional state is an abstraction of the
  participants focus of attention as their
  discourse unfolds.
• The state is a stack of focus spaces.
• A focus space (FS) is associated with a DS, and
  it contains DSP and objects, properties, and
  relations salient in the DS.
• When a DS ends, its FS is popped.
• When a DS starts, its FS is pushed onto the stack.

23
An example
DS1

• C1 I need to travel in May.
• A1 And, what day in May do you want
• to travel?
• C2 I need to be there for a meeting on 15th.
• A2 And you are flying into what city?
• C3 Seattle.
• A3 And what time would you like to
• leave Pittsburgh?
• C4 Hmm. I dont think there are many
• options for non-stop.
• A4 There are three non-stops today?
• C5 What are they?
• .

DS2
DS0
DS3
DS4
DS5
24
Discourse structure with intention info
DS0
DS1
DS3
DS4
DS2
DS5
A1-C2
A2-C3
C1
A3
C4-C7

• I0 C wants A to find a flight for C
• I1 C wants A to know that C is traveling in May.
• I2 A wants to know the departure data
• I3 A wants to know the destination
• I4 A wants to know the departure time
• I5 C wants A to find a nonstop flight

25
Problems with GS 1986

• Assume that discourses are task-oriented
• Assume there is a single, hierarchical structure
  shared by speaker and hearer
• Do people really build such structures when they
  speak? Do they use them in interpreting what
  others say?

26
Building discourse structure
27
Tasks

• Identify discourse segment boundaries
• Determine relations between segments
• Determine intentions of the segments
• Determine the attentional state
• Methods
• Inference-based approach symbolic
• Cue-based approach statistical

28
Inference-based approach

• Ex John hid Bills car keys. He was drunk.
• X is drunk ? people do not want X to drive
• People dont want X to drive ? people hide Xs
  car key.
• Abduction

? AI-complete Require and utilize world
knowledge.
29
Cue-based approach

• Attentional state
• Attentional changes
• (push) now, next, but, .
• (pop) anyway, in any case, now back to, ok,
  fine,...
• True interruption excuse me, I must interrupt
• Flashback oops, I forgot
• Intention
• Satisfaction-precedes first, second,
  furthermore, .
• Dominance for example, first, second, .

30
Cues (cont)

• Linguistic structure
• Elaboration for example,
• Concession although
• Condition if
• Sequence and, first, second.
• Contrast and,

31
One example

• (Marcu 1999) Train a parser on a discourse
  treebank.
• 90 trees, hand-annotated for rhetorical relations
  (RR)
• Learn to identify Elementary discourse units
  (EDUs)
• Learn to identify N, S, and their relation.
• Features WordNet-based similarity, lexical,
  structural,

32
Results

• Id EDUs 96-98 accuracy
• Id hierarchical structures (2 EDUs are related)
  Rec71, Prec84
• Id nucleus/satellite labels Rec58, Prec69
• Id rhetorical relation Rec38, Prec45
• ?Hierarchical structure is easier to id than
  rhetorical relations.

33
Why is Dialog Different?

• Why dialog is different
• Representing and interpreting dialog acts
• Dialogue structure and coherence

34
Dialogue and Conversational Agents

• What makes dialogue different?
• turn-taking
• grounding
• Implicature
• Speech act representation and interpretation.
• Approaches to coherence and structure.

35
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

36
Turns

• Conversation Analysis provides a socio-linguistic
  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.

37
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

38
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 utterance may occur across several turns
• A We've got 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

39
Review Dialogue is Different

• speakers know how to take turns know what?
  
• who should talk next, and when they should talk
• conversation analysis provides what?
• an approach to turn-taking
• transition-relevance places are typically at
  what place?
• utterance boundaries, but utterance
  segmentation is a difficult problem

40
Grounding

• Conversational participants must continually
  establish common ground (or mutual belief) H
  must ground S's utterances (by making it clear
  that believe understanding has occurred), or else
  indicate a grounding problem.
• Acknowledgement continuer / backchannel /
  acknowledgement token (also nods if vision
  available) to ground Ss utterance and to give
  back floor
• A returning on U.S. flight one.
• C Mm hmm
• Display (stronger method) display all or part of
  utterance to be grounded verbatim
• C OK I'll take the 5ish flight on the 11th.
• A On the 11th?
• Request for repair indicate lack of grounding
• C OK I'll take the 5ish flight on the 11th.
• A Huh?
• C I'll take the 5ish flight on the 11th.

41
Conversational Implicature

• Conversational Implicature is a particular class
  of licensed inference (that the speaker expects
  the hearer to draw).
• Grice's maxims for conversation explain how
  hearers draw such inferences.
• Example
• A What day in May did you want to travel?
• C I need to be there for a meeting that's from
  the 12th to the15th.
• A OK. There are 3 non-stops on the 11th.
• Implicature-licensed inferences
• the meeting information answers the request for
  travel dates
• there are not 4 non-stops

42
Grices Maxims

• Maxim of Quantity Be exactly as informative as
  is required
• Make your contribution as informative as is
  required (for the current purposes of the
  exchange).
• Do not make your contribution more informative
  than is required.
• Maxim of Quality Try to make your contribution
  one that is true
• Do not say what you believe to be false.
• Do not say that for which you lack adequate
  evidence.
• Maxim of Relevance Be relevant
• Maxim of Manner Be perspicuous
• Avoid obscurity of expression.
• Avoid ambiguity.
• Be brief (avoid unnecessary prolixity).
• Be orderly.
• Which maxims license previous inferences?

43
Dialogue Acts

• Austin (1962) observed that dialogue utterances
  are a kind of speaker action, or speech act.
• Example performative sentences
• I name this ship the Titanic.
• I second the motion.

44
Language Acts

• The utterance of any sentence in a real situation
  constitutes three kinds of act.
• Locutionary acts the utterance of a sentence
  with a particular meaning
• Illocutionary acts the act of asking, answering,
  promising, etc. in uttering a sentence
• Perlocutionary acts the (often intentional)
  production of certain effects upon the feelings,
  thoughts, or actions of the addressee in uttering
  a sentence.
• Example You cant do that. illocutionary
  force protesting, perlocutionary effect
  stopping or annoying the hearer

45
Speech Acts

• Searle uses term to describe illocutionary acts
  (1975).
• Assertives committing the speaker to something's
  being the case (suggesting, putting forward,
  boasting)
• Directives attempts by the speaker to get the
  addressee to do something (asking, ordering,
  requesting, inviting)
• Commissives committing the speaker to some
  future course of action (promising, planning,
  vowing, betting)
• Expressives expressing the psychological state
  of the speaker about a state of affairs
  (thanking, apologizing, welcoming, deploring)
• Declarations bringing about a different state of
  the world via the utterance (including
  performative acts I resign, you're fired)

46
Dialogue is different review 2

• Grounding (what is it?)
• a hearer must ground a speaker's utterances (by
  making it clear that (believed) understanding has
  occurred), or else indicate that a grounding
  problem occurred
• Implicature (what is it?)
• conversational implicature is a particular class
  of licensed inference (that the speaker expects
  the hearer to draw)
• Grice's maxims for conversation explain how
  hearers draw such inferences
• Dialogue Acts Austin (1962) observed that
  utterances are a kind of speaker action. (name
  the 3 acts?)
• The utterance of any sentence in a real situation
  constitutes three kinds of act locutionary,
  illocutionary, and perlocutionary.
• Searle uses the term speech acts for
  illocutionary acts (1975).

47
DAMSL

• A recent computational, expanded, hierarchical
  dialogue act tagging scheme (Dialogue Act Markup
  in Several Layers)
• Forward looking level (draws from Searle/Austin
  speech acts)
• Statement a claim made by the speaker
• Info-Request a question by the speaker
• Check a question for confirming information
• Influence-on-addressee Searle's directives
• Open-option a weak suggestion or listing of
  options
• Action-directive an actual command
• Influence-on-speaker Austin's commissives
• Offer speaker offers to do something (subject to
  confirmation)
• Commit speaker is committed to doing something
• Conventional other
• Opening greetings
• Closing farewells
• Thanking thanking and responding to thanks

48
DAMSL (cont)

• Backward looking level (draws from grounding,
  adjacency pairs, )
• Agreement speaker's response to previous
  proposal
• Accept
• Accept-part
• Maybe
• Reject-part
• Reject
• Hold
• Answer answering a question
• Understanding whether speaker understood
  previous
• Signal-non-understanding
• Signal-understanding
• Ack continuer or assessment
• Repeat-rephrase repetition or reformulation
• Completion collaborative completion

49
(No Transcript)
50
Dialogue Act Tagging Algorithms

• Sometimes there are obvious mappings from surface
  forms to dialogue acts
• STATEMENT I don't care about lunch.
• ACTION-DIRECTIVE Show me the flights from
  Pittsburgh.
• But there are also many violations, or Indirect
  Speech Acts
• ACTION-DIRECTIVE Can you show me the flights
  from Pittsburgh?
• ACTION-DIRECTIVE It's hot in here.
• A continuum of solutions
• plan inference model (derive only one of literal
  or indirect meaning)
• cue or idiom model (both literal and indirect
  meanings)

51
The Inferential Approach Searle

• Can you give me a list of the flights from
  Atlanta?
• X asked me whether I have the ability to give a
  list of flights.
• I assume X is being cooperative (in the Gricean
  sense) thus his utterance has some aim.
• X knows I am able to give a list, there is no
  reason why X should have a purely theoretical
  interest in my list-giving ability.
• Therefore X's utterance probably has some
  ulterior illocutionary point.
• A preparatory condition for a directive is that H
  have the ability to perform the action.
• Therefore, X has asked me about my preparedness
  for the action of giving a list.
• And, X and I are in a situation where giving
  lists is common expected.
• Thus, in the absence of another plausible
  illocutionary act, X is probably requesting I
  give him a list of flights.

52
Plan Inference / Recognition

• Making the inferential approach computational
• an AI planning (STRIPS) inspired model
  (preconditions, effects, body)
• Allen, Cohen, Perrault in the 70's, and others
  since
• Domain Acts
• BOOK-FLIGHT(A,C,F)
• Speech Acts
• INFORM(S,H,P)
• INFORMIF(S,H,P)
• REQUEST(S,H,ACT)
• Surface Acts
• SURFACE-REQUEST(S,H,ACT)

53
Plan Inference (cont)

• Plan Inference Heuristics
• Action-Effect Rule
• Precondition-Action Rule
• Body-Action Rule
• Know-Desire Rule
• Extended Inference Rule (prefix B(H,W(S)))
• See page 737 to trace indirect speech act
  interpretation of Can you give me a list of
  flights from Atlanta, e.g.
• S.REQUEST(S,H,InformIf(H,S,CanDo(H,Give(H,S,LIST))
  ))
• output REQUEST(S,H,Give(H,S,LIST))

54
Cue-Based Interpretation

• Less sophisticated, data-driven, more efficient
  alternative to plan inference.
• Multiple sources of knowledge provide dialogue
  act cues
• words and collocations
• prosody
• conversational structure
• combinations of the above

55
Words and Collocations

• please usually signals REQUEST
• word n-grams for each dialogue act (e.g., so you,
  sounds like are common REFORMULATION bigrams)

56
Prosody

• Decision trees for using prosody to classify
  speech acts
• (pg 742)
• Speech Acts (classes)
• STATEMENT (S)
• YES-NO QUESTIONS (QY)
• WH-QUESTIONS (QW)
• DECLARATIVE-QUESTIONS (QD)
• Prosody (features)
• pitch or fundamental frequency (F0) contour
• energy or loudness
• temporal duration

57
Conversational Structure

• Capture observations such as yeah is typically an
  AGREEMENT after a PROPOSAL but a BACKCHANNEL
  after an INFORM
• N-grams for dialogue act sequences
  (generalization of adjacency pairs)

58
Structure / Coherence in Dialogue

• Intentional (e.g. propose) versus an
  informational approach (e.g. causal
  relationships)
• Discourse structure a la Grosz and Sidner (1986)
• linguistic structure
• intentional structure
• attentional state
• Intentional structure
• discourse purpose
• discourse segment purpose (DSP)
• Two coherence relationships
• dominance DSP1 dominates DSP2 if satisfying
  DSP2 is intended to provide part of the
  satisfaction of DSP1
• satisfaction-precedence DSP1 satisfaction-preced
  es DSP2 if DSP1 must be satisfied before DSP2

59
Intentions

• Can be implemented using AI-planning formalisms
• Recognition algorithms range from inferential to
  cue-based
• Integration of Informational/Intentional
  coherence (Moore and Pollack (1992))

60
Wednesday

• Wednesday,
• More on Dialogue