The language of emotion, ethics and aesthetics - PowerPoint PPT Presentation

1 / 25
About This Presentation
Title:

The language of emotion, ethics and aesthetics

Description:

Appreciation (evaluation of phenomena) ... It was (Appreciation). June 22, 2006. Appraisal ... discriminating between affect, judgement and appreciation; ... – PowerPoint PPT presentation

Number of Views:109
Avg rating:3.0/5.0
Slides: 26
Provided by: jonath47
Category:

less

Transcript and Presenter's Notes

Title: The language of emotion, ethics and aesthetics


1
Appraisal Theory
  • The language of emotion, ethics and aesthetics

2
Introduction
  • Appraisal is concerned with the linguistic
    formulations of conveying emotions and opinions,
    how writers align their authorial personae with
    the stance of others, and how they manipulate
    their writings to convey a greater or lesser
    degree of strength and conviction in their
    propositions.

3
Presentation outline
  • Systemic Functional Linguistics
  • Appraisal resources
  • Attitude
  • Engagement
  • Graduation
  • Proposed research
  • Preliminary experiments
  • Discussion

4
Systemic Functional Linguistics
  • Systemic functional linguistics (Halliday 1994)
    identifies three metafunctions of language
    operating in parallel. Three levels of
    abstraction realise these types of meaning.

5
Systemic Functional Linguistics
  • Appraisal
  • elaborates on the notion of interpersonal
    meaning, describing how social relationships are
    negotiated through evaluations of self, others
    and artefacts.
  • is situated within discourse semantics
    attitude is often realised across grammatical
    boundaries, and is dynamic throughout a text.

6
Appraisal resources
appraisal
7
Attitude ways of feeling
  • Appraisal considers three types of attitude
  • Affect (personal emotion)
  • Judgement (appraisal of others behaviour) and
  • Appreciation (evaluation of phenomena).
  • All three ways of feeling can be either positive
    or negative.

8
Attitude ways of feeling
9
Attitude ways of feeling
  • Attitude can be realised
  • explicitly, through the lexicogrammar (inscribed)
    or
  • implicitly, through ideational meanings (invoked)
    that
  • are marked with attitudinal lexical items
    (flagged)
  • are elaborated by metaphor (provoked) or
  • make reference to cultural attitudinal norms
    (afforded)

10
Attitude ways of feeling
  • We smashed their way of life.
  • We brought the diseases. The alcohol.
  • We committed the murders.
  • We took the children from their mothers.
  • We fenced them in like sheep.
  • It was our ignorance and our prejudice.
  • And our failure to imagine these things being
    done to us.

FLAGGED
PROVOKED
AFFORDED
INSCRIBED
11
Engagement appraisals of appraisals
  • Engagement considers how writers convey their
    point of view and how they align themselves with
    respect to the position of others.
  • All utterances convey point of view (Stubbs 1996)
  • All utterances occur among a variety of other
    utterances regarding the same topic (Bakhtin
    1981)
  • Evaluative text is dialogic in that it responds
    to and anticipates the opinions of its audience
    and of other writers.

12
Engagement appraisals of appraisals
13
Graduation strength of evaluation
14
Computational Linguistics and Appraisal Theory
  • Taboada and Grieve (2004)
  • Determined a potential value of adjectives for
    affect, judgement and appreciation by calculating
    the PMI with the pronoun-copular pairs
  • I was (Affect)
  • He was (Judgement) and
  • It was (Appreciation).

15
Computational Linguistics and Appraisal Theory
  • Whitelaw et al. (2005)
  • Appraisal Groups are the atomic elements for
    sentiment analysis.
  • Attitude affect judgement appreciation
  • Orientation positive negative
  • Force low neutral high
  • Focus low neutral high
  • Polarity marked unmarked
  • Lexicon of frames constructed from WordNet, using
    the seed terms provided by Martin and White
    (2005).

16
Proposed research
  • Aim
  • Investigate techniques for the computational
    analysis of Appraisal
  • identifying appraisal-bearing propositions, and
  • labelling the constituents of the propositions
    according to the attitude, engagement and
    graduation typologies.

17
Proposed research
  • Motivation
  • Extend the breadth of sentiment analysis by
  • discriminating between affect, judgement and
    appreciation
  • determining author positioning with respect to
    others and
  • accounting for the vast range of linguistic
    features that serve to modify inscribed attitude
    and engagement.
  • Seek to statistically validate the Appraisal
    typologies.

18
Proposed research
  • Building an Appraisal Lexicon
  • The Appraisal typology might be populated
    semi-automatically by comparing the
    distributional similarity of target words with
    terms known to be appropriate to each node (as
    demonstrated by Weeds et al. (2005) for
    biomedical ontologies).

19
Proposed research
  • Finding patterns of Appraisal
  • The automatically acquired lexicon might be used
    to label sentences according to the type of
    attitude, engagement and graduation. Dependency
    parses of these sentences may reveal significant
    lexicogrammatical patterns of appraisal
    propositions.

20
Proposed research
  • Augmenting similarity measures with Appraisal
    annotations
  • The lexicogrammatical patterns of Appraisal might
    be used to provide additional features for
    semantic spaces (following Padó and Lapata
    (2004)). Can these features motivated from
    Appraisal Theory bootstrap the performance of
    metric in identifying words of attitude,
    engagement and graduation?

21
Proposed research
  • Challenges
  • Selecting seed terms
  • Some contexts confound the cultural norms
  • unpredictable behaviour vs. unpredictable plot
  • Infused graduation
  • disquieted, startled, frightened, terrified

22
Preliminary experiments
  • Similar words appear in similar contexts (Firth
    1957)
  • Can context determine Attitude?
  • Collect examples of Attitude
  • Cluster these examples using SenseClusters
    (Kulkarni and Pedersen 2005)

23
Preliminary experiments
  • Data collection
  • Martin and Whites (2005) example adjectives
  • Discarded 64 ambiguous terms
  • Discarded 151 terms with low probability mass
  • Extracted paragraphs from the BNC if they
    contained one of the 168 remaining adjectives

24
Preliminary experiments
  • Experiments with SenseClusters

25
Thank you!
  • David Hope drh21_at_sussex.ac.uk
  • Jonathon Read j.l.read_at_sussex.ac.uk
Write a Comment
User Comments (0)
About PowerShow.com