DA Week7 Lect1 by bzDalyw


       Models of
       Discourse Analysis
    Carolyn Penstein Rosé
 Language Technologies Institute/
Human-Computer Interaction Institute
Computational Approaches
                            Examples from the paper:
   Two steps
     Step  1: Metaphor
     Step 2: Metaphor
                            Lakoff’s concept:
   Does this paradigm      Metaphors structure how we think
                            about an event or state.
    cover everything that   The way we think affects:
                            (A) what we expect to happen,
    Lakoff and Johnson      (B) what we do,
                            (C) how we respond to what occurs
    place under the             during an event,
                            (D) and how we talk about what we
    heading of metaphor?        and others are doing
   Questions about presentations for next

   Rearranged syllabus slightly: see Drupal
   Posted responses to posts
   Readings for next unit + most of rest of semester posted
       Next unit focuses on Sentiment Analysis
            Product review dataset will be ready by next Monday for Assignment 3
       Note we won’t meet during Spring Break
   Unit 3 has a break too!
       We won’t meet on Wed, March 30 since several of us will be away
MIP: Metaphor Identification Procedure
Growing Interest?


Recent Approaches to Detection
   Peters and Peters 2000: Mined wordnet for abstract concepts that
    share word forms such as publication-publisher
   Mason 2004: Mine an internet corpus for domain specific selectional
    restriction differences
   Birke and Sarkar 2006: Start with seed sentences that have been
    annotated with figurative versus literal, and then do something like an
    instance based learning approach
   Gedigan et al. 2006: extract frames for MOTION and CURE from
    FrameNet, then extract sentences related to these from PropBank.
    Annotate by hand for metaphoricity. Use a maximum entropy
   Krishnakumaran and Zhu 2007: Look for sentences with “be” verb.
    Check for hyponymy using WordNet. If not there, look at bigram
    counts of subj-obj. If not high, then might be metaphorical.
What would Fass say?
   Problem with selectional restrictions as
     Will
         detect all kinds of nonliteral and
      anomalous language regardless if it is
      metaphorical or not
     Common metaphorical sense (i.e., “dead
      metaphors”) will fail here
     Some statements can be interpreted either
      way: “All men are animals”
Recent Approaches to Interpretation
   Metaphor based reasoning framework – reason in a
    source domain and apply reasoning to the target domain
    using a conceptual mapping
       Narayan’s KARMA 2004: parsed text as input
       Barnden and Lee’s ATT-Meta 2007: logical forms as input
   Talking Points 2008: uses WordNet, then uses minimal
    edits to bridge concepts
       Makeup is the Western burqa

   Shutova 2010: uses a statistical paraphrase approach
Shutova’s Take Away Message
   Approaches from the 80s and 90s were
    rule based
     Knowledge   engineering bottleneck
 Shutova’s work give some evidence that
  metaphor can be handled using a more
  contemporary (i.e., machine learning)
 Cast the metaphor interpretation problem
  as a paraphrase problem so you can use
  statistical machine translation approaches
Does paraphrase “cut it”?
Do you see a metaphor here?

 * How much of the problem can be solved by paraphrase?
Do you see metaphor here?
   Evey: Who are you?
    V: Who? Who is but the form following the function of what and what I
    am is a man in a mask.
    Evey: Well, I can see that.
    V: Of course you can, I’m not questioning your powers of observation,
    I’m merely remarking upon the paradox of asking a masked man who
    he is.
    Evey: Oh.
    V: But on this most auspicious of nights, permit me then, in lieu of the
    more commonplace soubriquet, to suggest the character of this
    dramatis persona.
   [pauses for a few seconds]
   Voila! In view humble vaudevillian veteran, cast vicariously as both
    victim and villain by the vicissitudes of fate. This visage, no mere
    veneer of vanity, is a vestige of the “vox populi” now vacant,
Data’s Identity
    We see evidence of how
     Data is framing his
    Do we see metaphor Note: The focus of the work of Shutova and
     here?               others who have self-identified as working on
                                        metaphor is on uncovering the literal
                                        meaning of expository text.
    Lakoff’s concept:
    Metaphors structure how we think
    about an event or state.
    The way we think affects:
    (A) what we expect to happen,
    (B) what we do,
    (C) how we respond to what occurs
        during an event,
    (D) and how we talk about what we
        and others are doing
Another spin on Metaphor
   Perspective modeling work
       Liberal versus Conservative
       Pro or Against
       Sentiment analysis more generally
   Different computational approach
       Skips step 1 – assumes all language represents perspective
       Simplifies step 2 – goal is to recognize a category rather than
   Usually models are based on word distributions
       Word vectors with weights
       Topic models
   We’ll explore this in the next unit
     Framing an Event in Progress
   Where does the
    paradigm for
    metaphors break
    down with
    examples like
      Step 1: recognize
      Step 2: map to
       literal meaning
   *** Still
    understanding a
    by comparison
    with another one
Breaking the Paradigm
   What can we do with conversational data?
     How  do we recognize that a metaphor is in
     What would it mean to do the interpretation?

To top