computational models of discourse analysis

28
Computational Models of Discourse Analysis Carolyn Penstein Rosé Language Technologies Institute/ Human-Computer Interaction Institute

Upload: aulii

Post on 05-Jan-2016

41 views

Category:

Documents


4 download

DESCRIPTION

Computational Models of Discourse Analysis. Carolyn Penstein Ros é Language Technologies Institute/ Human-Computer Interaction Institute. Warm-Up Discussion. There definitely is a content difference between male and female blogs in this corpus - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Computational Models of Discourse Analysis

Computational Models of Discourse Analysis

Carolyn Penstein Rosé

Language Technologies Institute/

Human-Computer Interaction Institute

Page 2: Computational Models of Discourse Analysis

Warm-Up Discussion There definitely is a content

difference between male and female blogs in this corpus

The question is whether there are also stylistic differences

Typical findings from genre analysis: women hedge more, use more indicators of involvement, and speak less formally These patterns are known to be

detectable through POS n-grams But is there something for

Engagement here? If women show more involvement,

wouldn’t you expect it to show up in Engagement like features?

What would you look for?

* What would you expect involvement to look like?

Analogy with educational research: Pretest score always accounts for most of the variance in posttest scores. If you don’t control for pretest score (by using it as a covariate in your comparisons of posttest) you frequently can’t see a meaningful difference between conditions. But when you do control for it, you frequently can. It almost always explains much less variance than pretest score. However, you can still see large effect sizes related to other factors.

Page 3: Computational Models of Discourse Analysis

Now read the excerpt from the Herring article

Does that change your view at all on the extent to which

Engagement is relevant?

Page 4: Computational Models of Discourse Analysis

Question about Assignment I also think that some sort of bootstrapping

may be useful for finding clusters of related words, and this time we have extra data that we could use. Would it be acceptable to use this data in the unlabelled form?

Yes!

Page 5: Computational Models of Discourse Analysis

Can you explain the difference between these three sentences:

They claimed that he is a stellar PhD student.

They stated that he is a stellar PhD student.

They demonstrated that he is a stellar PhD student.

Page 6: Computational Models of Discourse Analysis

How is “really” functioning in this passage, and what does that mean for Engagement oriented feature extraction?

Page 7: Computational Models of Discourse Analysis

Heteroglossia vs. Hedging

* What is your definition of a hedge? Is hedging the only function of Engagement?

Page 8: Computational Models of Discourse Analysis

Engagement Already established: Positioning a

propositionBut can it also be primarily positioning between

people?Patterns of positioning propositions as having

the same or different alignment between speaker and hearer could do this

Is positioning in communication always positioning by means of propositional content?

Page 9: Computational Models of Discourse Analysis

Gee vs. SFL Heteroglossia is like a tapestry

Gee was referring to the individual colored threads being woven together

The subtance of the perspective

Martin and White are referring to what holds the threads together

More focus on alignment versus disalignment

What happens when you borrow

other people’s words (Jim Gee’s

heteroglossia) but present them

without markers of alignment or

disalignment (SFL monoglossia)?

Page 10: Computational Models of Discourse Analysis

Connection between Heteroglossia and Attitude

But is this really different from a disclaim?

And is this really different from a proclaim?

Page 11: Computational Models of Discourse Analysis

Student Comment I agree with what David said in class on

Monday--that heterogloss, as defined in the reading, sounds much more like hedging than what Gee's and what I imagine Bakhtin's concepts of it were. And as such, I believe hedging is a much more effective tool in showing generational or occupational differences rather than gender differences.

Page 12: Computational Models of Discourse Analysis

Accommodation

by pointing out the inflation of Saddam’s body count by neocons in an effort to further vilify him and thus further justify our invasion we are not DEFENDING saddam....just pointing out how neocons rarely let facts get in the way of a good war.

by pointing out the inflation of Saddam’s body count by neocons in an effort to further vilify him and thus further justify our invasion we are not DEFENDING saddam....just pointing out how neocons rarely let facts get in the way of a good war.

So wait, how many do you think Saddam killed or oppressed? You’re trying to make him look better than he actually was. You’re the one inflating the casualties we’ve caused! Seriously, what estimates (with a link) are there that we’ve killed over 100,000 civilians. Not some crack pot geocities page either.

So wait, how many do you think Saddam killed or oppressed? You’re trying to make him look better than he actually was. You’re the one inflating the casualties we’ve caused! Seriously, what estimates (with a link) are there that we’ve killed over 100,000 civilians. Not some crack pot geocities page either.

Page 13: Computational Models of Discourse Analysis

Other views of positioning From Tannen’s

Framing in Discourse Words that sound like a

role The substance of that

role – in the Gee sense

In this case, speaking the words shows alignment because you use them as though they are your own

Page 14: Computational Models of Discourse Analysis

Connections Alignment and Solidarity (positioning

between people)Can you think of other indicators of solidarity

Personal pronouns Accommodation Friendly language

Factivity (positioning between people and propositions) I regret eating that chocolate Is this heteroglossic?

Page 15: Computational Models of Discourse Analysis

How would an Engagement style analysis compare/relate to the LIWC analysis we discussed last time?

Page 16: Computational Models of Discourse Analysis

Student Comment Iris: I think that not only may it sound sexist

to hypothesize about what people of a certain gender talk like but also may the way people talk be more indicative of where they are from.

Page 17: Computational Models of Discourse Analysis

How would you expect an Engagement style analysis to relate to personality?

•What effect would you expect to see on conversations?

•Are these necessarily connected?

Page 18: Computational Models of Discourse Analysis

Freshman Engineering Study 131 Freshman engineering students

worked in groups of 3 or 4 to design a better wrench Applying principles related to stress

and leverage

Procedure Tutorial on computer aided engineering Pretest Collaborative design activity Posttest Questionnaire

Page 19: Computational Models of Discourse Analysis

Heteroglossia Manipulation

Page 20: Computational Models of Discourse Analysis

Social Manipulation

Page 21: Computational Models of Discourse Analysis

Tutor Agent Design

RequestDetector

T.T

akin

gC

oord

inato

r

TutoringManager

Outp

utC

oord

inato

r

ConcertChatActor

ConcertChatListener

ConcertChat Server

ConcertChat Server

TutoringActorPromptingActor

MessageFilterDiscourseMemory PresenceFilter

PromptingManager

IntroductionsActor

IntroductionsManager

ActivityDetector

AnnotationFilter

PlanExecutor

SocialController

Pro

gre

ssD

ete

ctor

TutorOne last thing on this topic, Does more (or less) stress in a wrench make it easier to use?

S95 no change?

TutorYou are correct about that.. Stress doesn't determine ease of use.

TutorIt's the moment achieved by the wrench that determines the ease of use.

S89 yay!

TutorIt's good to have your contributions in the discussion Jackie :-)

Tutor Go team :-)S89 Go team yay

TutorI am happy to be working with our team

S89 Me tooS95 whoa the bot knows my name

TutorUnfortunately maximum stress (12800) in our design1 is way above the maximum allowed stress (i.e. 8750)

TutorThis wrench cannot be safely used!

Kumar, R. & Rosé, C. P. (2011). Architecture for building Conversational Agents that support Collaborative Learning, IEEE Transactions on Learning Technologies special issue on Intelligent and Innovative Support Systems for Computer Supported Collaborative Learning

Page 22: Computational Models of Discourse Analysis

Results on Breadth of Coverage of Design Space

Significant main effect of Heteroglossia on number of ideas mentionedHeteroglossia was better than

Monoglossia and Neutral Significant interaction

In the Social condition, Monoglossia was worse than the other two

Page 23: Computational Models of Discourse Analysis

Results on Perception Students were significantly happier with the interaction in

the Heteroglossia condition than Neutral, with Monoglossia in the middle

Students liked the Heteroglossic and Monoglossic agents better than the Neutral agent

Students in the Heteroglossia condition felt marginally more successful than students in the Monoglossia condition

No effect on Personality indicators such as Pushy, Wishy Washy, etc.

Does that mean that impression of personality and how you feel about an interaction with someone are not linked?

Page 24: Computational Models of Discourse Analysis

Student Comment I would also note that English is a very

gender neutral language, so gender performativity is harder to classify.

Page 25: Computational Models of Discourse Analysis

Discussion from Last Time How would you achieve a good balance

between an operationalization of Engagement that is useful and yet attainable in terms of reliability of coding?

Page 26: Computational Models of Discourse Analysis

Hedging and Occupation? And as such, I believe hedging is a much

more effective tool in showing generational or occupational differences rather than gender differences. For example, teenagers often use verbs such

as 'like' and 'all' to report speech: he was all 'that's stupid' and then he was like ''but I'm stupid too'. The occupational differences I would attribute to the differences between people who need exact values as opposed to people who can accept generalizations or approximations.

Page 27: Computational Models of Discourse Analysis
Page 28: Computational Models of Discourse Analysis

Questions?