Docstoc

Overview TIME IS MONEY… Conceptual Metaphor Conceptual

Document Sample
Overview TIME IS MONEY… Conceptual Metaphor Conceptual Powered By Docstoc
					                                                                   Overview
                                                                   Part I:
                                                                       Linguistic theories of metaphor
                                                                       Psycholinguistic processing models and experiments

               ESSLLI-2007 | Figurative Language Processing
                                     Birte Lönneker-Rodman
                                                                   Part II: Metaphor and Computation
                    International Computer Science Institute
                                          Berkeley, CA, USA




                                                                                     ESSLLI-2007 Figurative Language Processing                2




TIME IS MONEY…                                                     Conceptual Metaphor
…so let’s get started!                                               A conceptual domain commonly understood in terms of
                                                                     another
Examples:                                                                                                 can be understood in
                                                                                      TIME                                         MONEY
                                                                                                                terms of
                                                                      standard
                                                                                      TIME                               IS        MONEY
That flat tire cost me an hour.                                       notation

You need to budget your time.                                         general
                                                                                   conceptual                                     conceptual
                                                                      notation                                           IS
Don't spend too (much|little) time on a slide.                                      domain A                                       domain B
                                                                      pattern
I lost a lot of time when I got sick.                                             target domain
                                                                                                is understood in terms source domain
                                                                    terminology       (more
                                                                                                          of           (less abstract)
                                                                                     abstract)
                ESSLLI-2007 Figurative Language Processing     3                     ESSLLI-2007 Figurative Language Processing                4




                                                                   Linguistic manifestation of
Conceptual domain                                                  conceptual metaphors
  “any coherent organization of experience” (Kövecses              Some commonly used lexical units to talk about target
  2002:4)                                                            domain (e.g. TIME) are “borrowed from” source domain
  cluster of semantically and/or encyclopedically related            (e.g. MONEY)
  concepts or “constituent elements”                                 ⇒metaphorically motivated polysemy (e.g. of save, cost,

  Several, but not all source                                          spend)
  domain constituents (and
  relations) play a role in                                        Terminology
  target domain                                                     “metaphorical linguistic expressions” (Kövecses 2002:4)
                                                                    or
   ⇒systematic but partial mapping                                  lexical metaphors
                ESSLLI-2007 Figurative Language Processing     5                     ESSLLI-2007 Figurative Language Processing                6
LIFE IS A JOURNEY and
LOVE IS A JOURNEY                                                                 Summary
Examples containing lexical metaphors:                                            Conceptual metaphor         conventionalized mapping between two different
   1.    Look how far we've come.                                                                             domains of experience
   2.    He has never let anyone get in his way.                                  Lexical metaphor            linguistic expression illustrating a conceptual
   3.    We'll just have to go our separate ways.                                                             metaphor
   4.    Our marriage is on the rocks.                                            Systematic mapping          several lexical metaphors illustrate the same
   5.    She has gone through a lot in life.                                                                  conceptual metaphor
   6.    We are at a crossroads.                                                  Partial mapping             not every entity from the source domain gives rise to a
                                                                                                              lexical metaphor
Exercise/questions:                                                               Highlighting effect         entities participating in the mapping highlight particular
                                                                                                              (real or imagined) aspects about target domain
  Can you identify the appropriate target domain illustrated by each
                                                                                  Many-to-many                  one source domain can map onto several different
  sentence?                                                                       relationship                    targets
  Which conceptual entities from the JOURNEY source domain are                                                  one target domain can be understood in terms of
  mapped? Which effect does this have?                                                                            several different sources (e.g., think of other
                                                                                                                  source domains for LOVE)
                    ESSLLI-2007 Figurative Language Processing                7                         ESSLLI-2007 Figurative Language Processing                                8




Metaphor Theories                                                                 Metaphor Theories
Theory of Conceptual Metaphor:
        groundbreaking work by George Lakoff and Mark Johnson
        (1980)
        several earlier accounts (e.g. Michael Reddy on IDEAS
        ARE OBJECTS in 1979)
                                                                                            Dew             Veil                                                   ad-hoc class
        further developments include Conceptual Blending theory                                                                                      SITUATION
                                                                                        liquid   covering         solid                               unpleasant
        (Fauconnier and Turner 2002)                                                 amorphous transparent      textured                               confining
                                                                                      colorless shimmering       colorful                            unrewarding       is-member-of
                                                                                          …         …              …                                      …
                                                                                                                                  my job                                jail

                    ESSLLI-2007 Figurative Language Processing                9                         ESSLLI-2007 Figurative Language Processing                             10




                                                                                  Rich images vs. conceptual
Rich image metaphors                                                              metaphors
   1.    Dew is a veil. ‘something that covers partly and is shimmering,                                                 Rich image Conceptual metaphor
         like a veil’                                                             Level of abstraction                  low         high
   2.    This surgeon is a butcher. ‘someone who uses coarse methods,             Systematicity of mapping              absent/low  high
         like a butcher, for a job that requires finesse and involves a
         knife as an instrument’                                                  Role of cultural
                                                                                                                        more
         We haven't time to give it more than a catlick. [BNC]                    knowledge for                                                  less important
   3.                                                                                                                   important
         ‘perfunctory wash, similar to that done by cats licking their fur’       comprehension
   4.    Despite this the market treats Evans like a black sheep. [BNC]                                                 from                     rather conventional, but
         ‘someone who is considered embarrassing, less successful or              Conventionality                       established              novel extensions
         more immoral than the rest of the group’                                                                       to novel                 possible
   5.    My horse with a mane made of short rainbows. (Navaho song
         cited by Lakoff 1993:230)                                                (cf. e.g. Lakoff 1987:444–456; Dobrovol'skij and Piirainen
                                                                                    2005:161–165)
                    ESSLLI-2007 Figurative Language Processing            11                            ESSLLI-2007 Figurative Language Processing                             12
Main types of                                                                   Main types of
novel metaphor (I)                                                              novel metaphor (II)
1. Lexical extension of conventional conceptual metaphors                       2. Analogies

Extended set of mapped lexical items from the source domain                     make use of several well-established conceptual mappings

THEORIES ARE CONSTRUCTED OBJECTS                                                   Example: My job is a jail.
   Example 1: He is trying to buttress his argument with a lot of
                                                                                   Understanding involves independently existing conceptual
     irrelevant facts, but it is still so shaky that it will easily fall             metaphors:
     apart under criticism.                                                        PSYCHOLOGICAL FORCE IS PHYSICAL FORCE
     (conventional lexical metaphors)                                              ACTIONS ARE SELF-PROPELLED MOVEMENTS
   Example 2: Your theory is constructed out of cheap stucco.
     (one conventional and one novel lexical metaphor)                          3. Rich image metaphors

                   ESSLLI-2007 Figurative Language Processing              13                     ESSLLI-2007 Figurative Language Processing       14




                                                                                Potential of Conceptual Metaphor for NLP
Summary                                                                         and reasoning
  Different theories have been proposed for metaphor                              systematic and economical representation of metaphorically
  The conceptual theory of metaphor                                               motivated polysemy
      deals with conventionalized metaphorical word senses                        (e.g. guidelines for sense splitting)
      illustrating mappings between conceptual domains                            infer sense of novel lexical metaphors that extend a
                                                                                  conventional metaphorical mapping
  Different strategies for novel metaphors
      e.g. extend set of lexical metaphors illustrating a domain                  creatively generate such novel metaphors
      mapping                                                                     use inferences from conceptual metaphors to understand and
  Finding out how humans process metaphor is difficult                            generate analogies
      role of conceptual domain mapping in this process still subject to          reuse inferences from source domain to compute information
      research                                                                    about target domain, including speaker’s/writer’s attitude or
                                                                                  discourse goal

                   ESSLLI-2007 Figurative Language Processing              15                     ESSLLI-2007 Figurative Language Processing       16




Overview of Part II:
Metaphor and Computation                                                        Selectional restrictions
                                                                                    ‡ ‡š‹        ‡†‰‡ „ƒ•‡
                                                                                  •‹ ’Ž Ž …ƒŽ‘™ Ž
Metaphor and selectional restrictions                                           ‡”„•
Metaphor reasoning systems                                                         valence (arity)         selectional restrictions on arguments
                                                                                     digest(x, y)             animate(x)              edible(y)
Metaphor resources (databases, annotated corpora)
                                                                                    devour(x, y)              animate(x)              edible(y)
Metaphor detection
                                                                                 ‘—•
                                                                                   animate(child)             ¬edible(stone)
                                                                                   animate(professor)         ¬edible(fact)
                                                                                   edible(spinach)            ¬edible(book)
                                                                                   edible(steak)

                   ESSLLI-2007 Figurative Language Processing              17                     ESSLLI-2007 Figurative Language Processing       18
Selectional restrictions and                                                                              Solutions to violation of selectional
metaphor                                                                                                  restrictions
Sentences licensed by lexical knowledge base:                                                             Example 1: The variable N goes from 1 to 100.
       The child                 devoured                              the spinach.                        (first argument of go should be animate)
       The professor             was still digesting                   the steak.
              animate                                                       edible
                                                                                                          Solution: allow variables to be at a certain location and
                                                                                                           define go as “change location”
Sentences not licensed by lexical knowledge base:
                                                                                                              ⇒release selectional restriction of first argument of go:
It is difficult for    me               to digest          all these facts.
                                                                                                                “things that can be located in space”
                       She              devoured the book                           in one day.                  ⇒   more general meaning of go
                        animate                                      ¬edible                                  ⇒hard-code conventional metaphor STATES ARE
                                                                                                                LOCATIONS with respect to variables
Conceptual metaphor: IDEAS ARE FOOD
                        ESSLLI-2007 Figurative Language Processing                                   19                          ESSLLI-2007 Figurative Language Processing                                 20




Solutions to violation of selectional                                                                     Lexical metaphors as different word
restrictions                                                                                              senses? WordNet 3.0
                                                                                                          Polysemy of go                                                  Polysemy of elephant
Example 2: Mary is graceful, but John is an elephant.                                                     1.{travel, go, move, locomote} (change location;                1.elephant(five-toed pachyderm)
  (first argument of be an elephant should be an elephant)                                                move, travel, or proceed, also metaphorically)                  2.elephant(the symbol of the
                                                                                                                  We travelled from Rome to Naples by bus                 Republican Party; introduced in
Solution: define that if something is clumsy and some other                                                       The policemen went from door to door looking            cartoons by Thomas Nast in
                                                                                                                       for the suspect                                    1874)
  unspecified properties hold, it is an elephant; by a                                                            news travelled fast
                                                                                                                        [but: ?news moved fast, ??news locomoted
  reasoning process called abduction, infer that what is                                                                     fast]
  meant is ‘John is clumsy’                                                                               2.{run, go, pass, lead, extend} (stretch out over a
                                                                                                          distance, space, time, or scope; run or extend
    ⇒ different meaning of be an elephant is licensed (be clumsy)                                         between two points or beyond a certain point)
    ⇒ this is the most likely interpretation of the sentence (good                                                Service runs all the way to Cranbury
       contrast to graceful)                                                                                      His knowledge doesn't go very far
                                                                                                                  [cf. The variable N goes from 1 to 100.]
Hobbs (1992)                                                                                              3.[...]



                        ESSLLI-2007 Figurative Language Processing                                   21                          ESSLLI-2007 Figurative Language Processing                                 22




Sense inventories vs.                                                                                     Metaphor reasoning systems:
metaphor reasoning                                                                                        Overview
                        Sense inventories                                   Reasoning
                WordNet                                       TACITUS system (interpretation as
                                                                                                            knowledge-based
Example
                                                              abduction, Hobbs 1992)
                  “all” senses                                  one sense/a subset of senses;
                                                                                                            exploit knowledge of systematic language
Available
information
                                                                (at least partial) information on           conventions (conceptual metaphor, mapping)
                                                                  possible mapping(s)
                Fixed:                                        flexible                                        ⇒avoid resorting to more computationally expensive methods
                - reference for comparison and                - recognition of more novel and
Fixedness
                competition (e.g. WSD-exercises);             creative senses; but difficult to define      reason in source domain and transfer results to target
versus
                but novel senses not covered                  and compare                                   domain
flexibility
                - readily available for annotation of         - inferences give access to additional
                running texts                                 information such as speaker attitude
                                                                                                            examples: TACITUS, KARMA, ATT-Meta



                        ESSLLI-2007 Figurative Language Processing                                   23                          ESSLLI-2007 Figurative Language Processing                                 24
Metaphor reasoning in KARMA and                                                Components of metaphor reasoning
ATT-Meta                                                                       systems
KARMA: a story understanding system (Narayanan 1999)
    produces time-sliced representations of development in target
    domain (Economy)
    derived from event phrases such as European Giant falls sick
    represents resulting (end) state in terms of feature structures

ATT-Meta: a Question-Answering system (Barnden et al. 2002)
    is given a possibly metaphorical representation of current state of
    the world
    verifies whether a particular fact (submitted by user as a
    “question”) holds


                 ESSLLI-2007 Figurative Language Processing               25                 ESSLLI-2007 Figurative Language Processing   26




Components of metaphor reasoning                                               Components of metaphor reasoning
systems                                                                        systems




                 ESSLLI-2007 Figurative Language Processing               27                 ESSLLI-2007 Figurative Language Processing   28




Metaphor reasoning systems                                                     Mappings in KARMA
  extensive domain knowledge                                                   Ontological maps for entities and objects
    hand-coded
  computational engines                                                          MOVER ─ ACTOR
    derive new facts or consequences within domains (usually source            Schema maps for events, actions, and processes
    domain)
    design and implementation: neural networks or formal logic                   move ─ act
  Stochastic Petri Nets and Belief networks in KARMA                             fall ─ fail
  situation-based and episode-based first-order logic in ATT-                  Parameter maps for features of objects or events
  Meta
  mapping rules allow information flow from one domain to the                    velocity ─ rate of progress made
  other                                                                          distance traveled ─ degree of plan completion
    central components, though usually simpler than domain
    representations
                 ESSLLI-2007 Figurative Language Processing               29                 ESSLLI-2007 Figurative Language Processing   30
An ATT-Meta Example                                                    Mappings in ATT-Meta
“Metaphors of Mind”                                                     Encode only conventional knowledge on conceptual
                                                                        metaphors (“minimal mappings”, partial mappings)
MIND AS PHYSICAL SPACE/CONTAINER
    keep an idea in mind, get an idea out of your mind, to have          ⇒no explicit mappings for novel extensions of conceptual
    a lot on your mind, (an idea) comes to your mind, ...                  metaphors
                                                                        When novel lexical metaphor is encountered, reason
IDEAS ARE PHYSICAL OBJECTS
    hide, give away, chew, digest an idea (physical operations          in “pretence cocoon” where utterance is taken as
    on ideas)                                                           literally true (source domain)
                                                                         ⇒derive “pretence” facts for which conventional mappings
Example to be reasoned upon:                                               exist
    In the far reaches of her mind, Anne knew Kyle was
    having an affair [...].
                ESSLLI-2007 Figurative Language Processing        31                     ESSLLI-2007 Figurative Language Processing                    32




Metaphor Reasoning Systems:
Problems                                                               Metaphor databases
  Acquire sufficient and suitable knowledge                             The Berkeley Master Metaphor List
  ⇒transform knowledge from lexical databases?                             ca. 1,700 sentences or phrases listed under mapping headlines
                                                                              They are at a crossroads in their relationship
  ⇒mine corpora?
                                                                              listed under LOVE IS A JOURNEY
  Provide suitable input representations for these                      ATT-Meta Project Databank of Metaphors of Mind
  systems                                                                  1,070 categorized text examples and 65 speech examples
  ⇒build appropriate parsers for free natural language texts                 (* ONCE I'D LET THAT LITTLE PIECE OF SELF-KNOWLEDGE
                                                                             FLOAT TO THE TOP OF MY MIND, a terrible rage began to seize
  Interpret output representations from these systems                        hold of me.*)
                                                                             listed under mapping MIND AS PHYSICAL SPACE
  ⇒generate natural language text from output feature structures
    or logical forms                                                    The Hamburg Metaphor Database
                                                                           1,483 French and 173 German sentences or phrases annotated with
                                                                           domain mappings and EuroWordNet synsets for the lexical metaphor

                ESSLLI-2007 Figurative Language Processing        33                     ESSLLI-2007 Figurative Language Processing                    34




Hamburg Metaphor Database:
Example                                                                Metaphor Annotation
  Source domain: A different path;                                      Results reported for metaphorical – non metaphorical
  Target Domain: A different means of achieving the purpose;            distinction
  Language: French
                                                                        Annotators trained in Conceptual Metaphor theory
                                                                         ⇒ Conventional lexical metaphors count as metaphorical

                                                                                                                                Kappa tests
                                                                                                                        (reproducibility of results)
                                                                              Lee (2006)                                     from 0.82 to 0.85
                                                                        Lönneker-Rodman (2007)                            0.897 (Cohen's kappa)
                                                                         ⇒ relatively stable inter-annotator agreement compared to other
                                                                           semantic annotation tasks

                ESSLLI-2007 Figurative Language Processing        35                     ESSLLI-2007 Figurative Language Processing                    36
Automatic Metaphor Detection                                                                  A clustering approach (I)
Focusing on verbs as lexical metaphors:                                                       Birke and Sarkar (2006)
      Mason (2004) – detection of domain mappings                                              “literal vs. nonliteral language”
      Birke and Sarkar (2006) – literal vs. nonliteral language in                             no reference theory, (annotator) decision mainly based
      general                                                                                  on selectional restrictions and “world knowledge”
      Gedigian, Bryant, Narayanan and Ciric (2006)
                                                                                               most examples are metaphors
                                                                                                     Literal: The girl and her brother grasped their mother’s hand.
Including also comparison/categorization metaphors:                                                  Nonliteral: He thinks he has grasped the essentials of the
      Krishnakumaran and Zhu (2007) – distinguish between                                            institute’s finance philosophies.
      conventional and novel metaphors in the Master Metaphor
      List
                     ESSLLI-2007 Figurative Language Processing                          37                           ESSLLI-2007 Figurative Language Processing                        38




A clustering approach (II)                                                                    A clustering approach (III) – Data
Algorithm overview                                                                            sets
  modification of WSD algorithm by Karov                                                      Test set: sentences from the Wall Street Journal (WSJ)
  and Edelman (1998)                                                                          Nonliteral feedback set:
  calculate similarities between sentences containing target word                               example sentences from list of idioms, sayings and slang and Master
  (lexical metaphor) and literal and nonliteral sentence                                        Metaphor List
  collections (feedback sets) for each verb                                                     WSJ sentences containing target verb or a synonym of this nonliteral
                                                                                                meaning (e.g. grasp, comprehend); synonym found in the above sources
      original algorithm: attract test sentence to collection containing the most
      similar sentence                                                                        Literal feedback set:
      modified version: calculate similarities to all sentences in collection and                WSJ sentences containing target word or synonym from same WordNet
      attract test sentence to the collection with the largest similarity sum                    synset (e.g. grasp, grip)
  two sentences are similar if they contain similar words and two                                example sentences for these synsets from WordNet itself
  words are similar if they are contained in similar sentences                                   noisy due to possibly nonliteral meaning of synset
                                                                                                 ⇒   subsequent automatic “cleaning” (remove sentence or move it to nonliteral set)
      features: stemmed verbs and nouns in the sentence and surrounding                                 relies on presence of phrasal verbs (e.g. step down) and on stem overlap with
      sentences                                                                                         nonliteral set

                     ESSLLI-2007 Figurative Language Processing                          39                           ESSLLI-2007 Figurative Language Processing                        40




A clustering approach (IV) – Results                                                          Maximum entropy classification (I)
Test set                                                                                      Gedigian et al. (2006)
  1,298 sentences illustrating usages of 25 verbs (1 to 115 occurrences each)
  manually annotated                                                                          Theoretical background: Cognitive Theory of Metaphor
      Cohen’s kappa for two annotators on 200 examples: 0.77                                  Verbs to be disambiguated: from FrameNet frames
                                                                                                     Motion-related frames (Motion, Motion-directional, Self-motion, Cause-
                                original KE modified              modified algorithm                 motion); examples: glide, go; drop, fall; crawl, hobble; catapult, throw, ...
                                 algorithm algorithm              with active learning               Placing frame; examples: cram, heap, tuck, ...
        F-score                                                                                      Cure frame; examples: cure, heal, treat, ...
(2 x precision x recall)/         36.90%              53.80%           64.90%                 Labeled training data: 4,186 occurrences of these verbs in Wall Street
   precision + recall                                                                           Journal corpus
  F-score calculated based on average precision and average recall                                   labels: metaphor, literal, unclear; only clear annotations used for training
      Average precision: average of literal and nonliteral precision                                 more than 90% of data are lexical metaphors
  Precision and recall not reported individually                                              Question: How can we explain this high metaphor ratio?
                     ESSLLI-2007 Figurative Language Processing                          41                           ESSLLI-2007 Figurative Language Processing                        42
Maximum entropy classification (II):                                                             Maximum entropy classification (III):
Features                                                                                         Results
 Verb bias                                                                                       Selection of most successful feature combination on validation
       most verbs show clear tendency towards literal or metaphorical uses                         set
 Semantic type of verb arguments                                                                      verb bias
 1. Extract argument information from PropBank annotation of WSJ corpus
       example: argument 1 typically realized as direct object in active English sentence
                                                                                                      ARG1_TYPE
          e.g. Traders threw stocks out of the windows
 2. Extract head word of each argument; e.g. out of the windows                                  Results with these features
 3. Determine “semantic type” of head word                                                            Test set: 861 occurrences
       pronoun: human/non-human/ambiguous
       Named Entity: NE tag assigned by NE recognizer                                                          Majority baseline            Majority baseline per   Classifier
       other noun: WordNet synset containing head word                                                             overall                           verb             result
 Example: [The drugARG3] is being used primarily to treat [anemiasARG2].                          Accuracy         92.90%                          94.89%           95.12%
   ARG2_TYPE={anemia}
   ARG3_TYPE={drug}                                                                              No measures other than accuracy are given.

                        ESSLLI-2007 Figurative Language Processing                          43                      ESSLLI-2007 Figurative Language Processing                   44




Summary                                                                                          Hidden Slides
 We have seen…
  Metaphor as selectional restriction
  Metaphor reasoning systems with hand-crafted rules and
  knowledge bases
  Metaphor databases with manual annotations
  Metaphor detection algorithms for collecting lexical
  metaphors automatically
       Need annotated/pre-classified data sets
       Focus on verbs

                        ESSLLI-2007 Figurative Language Processing                          45                      ESSLLI-2007 Figurative Language Processing                   46




How do humans process metaphors?
                                                                                                 How do humans process
Models and Predictions                                                                           metaphors? Experiment I
 Standard pragmatic model (Grice 1975; Searle 1979)                                              Test question: Is this sentence meaningful?
    Sequentiality: literal meaning attempted first, a conflict is detected, finally
      metaphorical meaning is found                                                                 Example 1: Some mouths are sewers. (figurative: novel image
    ⇒ longer processing times for metaphor than literal language                                      metaphor)
                                                                                                    Example 2: Some tunnels are sewers. (literal)
 Cognitive linguistics model (Lakoff 1993)                                                          Example 3: Some lamps are sewers. (nonsense)
    Conceptual metaphors used with no noticeable effort                                               same processing times for metaphors and literal language
    ⇒ same effort and processing times for metaphor and literal language                              but: difficulty of designing the test (semantic plausibility, word
                                                                                                      frequency, ...)
 Current psycholinguistic processing models (e.g. Gibbs 1994;
   Glucksberg et al. 1997)                                                                       (McElree and Nordlie 1999)
    Note correspondences between domains (alignment) and selectively
      project properties (cf. features, but also highlighting)
    ⇒ more effort for processing metaphor than literal language

                        ESSLLI-2007 Figurative Language Processing                          47                      ESSLLI-2007 Figurative Language Processing                   48
How do humans process                                                                                      Conceptual metaphors
metaphors? Experiment II                                                                                   are reality
Measurement of event-related brain potentials (ERPs)                                                          1.   Read story about anger (e.g. someone returns John’s car with new
                                                                                                                   dents in it)
                                                                                                              2.   Read priming context (one of the following):
Test sentences                                                                                                      Sentence 1: He blew his stack. (idiom)
  Example 1: That stone we saw in the natural history museum is a                                                   Sentence 2: He got very angry. (literal)
  gem. (literal)                                                                                                    Sentence 3: He saw many dents. (neutral control phrase)
  Example 2: After giving it some thought, I realized the new idea was                                        3.   Do lexical decision task: “Is this string of letters an English word?”
                                                                                                                      heat
  a gem. (relatively novel metaphor extending IDEAS ARE                                                               (an unrelated word)
  OBJECTS)                                                                                                            (a nonword)
      comprehension of metaphorical sentences more difficult than that of                                     Subjects primed by Sentence 1 responded faster for heat, but not for other
      literal sentences
                                                                                                                  words.
      comparable with difference between high- and low-frequency words
                                                                                                              ⇒ Sentence 1 evokes the mapping ANGER IS HEAT
      still, literal and metaphoric language share some processing
      mechanisms
(Coulson and van Petten 2002)                                                                                 (Gibbs 1994)
                          ESSLLI-2007 Figurative Language Processing                                  49                        ESSLLI-2007 Figurative Language Processing                  50




Conceptual metaphors might not be                                                                          MIND IS PHYSICAL SPACE
“embodied” reality                                                                                         conversion rules (ATT-Meta)
Do metaphors rely on embodied representations from source domain?                                          IF is-idea(J) AND is-person(P)
       Example: grasping an idea might go back to grasping a spoon                                         AND WITHIN PRETENCE: is-physical-region(mind-of(P))
                                                                                                           AND WITHIN PRETENCE: physically-in(J, mind-of(P))
Monitor brain activity with functional magnetic resonance imaging                                          THEN {presumed} to-degree-at-least(very-low):
 (fMRI)                                                                                                    can-mentally-operate-on(P, J).
       Results contradict this embodiment; but inconclusive

Methodological issues:                                                                                     Source domain reasoning: The far reaches of a space are a physical
    1. Difficulty of visually representing grasping an idea for the test                                     part of that space, so what is in the far reaches of your mind is
    2. Understanding of conventional mapping, once learned, might no longer                                  physically “in” your mind.
       rely on motor representations                                                                       Apply conversion rule: You can mentally operate on an idea that is in
                                                                                                             the far reaches of your mind, to a very low degree at least.
(Aziz-Zadeh et al. 2006)

                          ESSLLI-2007 Figurative Language Processing                                  51                        ESSLLI-2007 Figurative Language Processing                  52




IDEAS ARE PHYSICAL OBJECTS conversion
rules (ATT-Meta)
IF is-idea(J) AND is-person(P)
AND WITHIN PRETENCE: is-physical-object(J)
AND WITHIN PRETENCE: is-person(conscious-self-of(P))
AND WITHIN PRETENCE: to-degree-at-least(Degree):
can-physically-operate-on(conscious-self-of(P), J)
THEN {presumed} to-degree-at-least(Degree):
can-consciously-mentally-operate-on(P, J).

Source domain reasoning: something that is far away is accessible to physical
  operation only to a very low degree;
  the far reaches of something are far away from the center of that thing, so the degree
  of possible physical operation on an object in the far reaches is very low
Apply conversion rule: default degree (very-low) from previous rule is not increased
    ⇒ Anne had a very low degree of ability to consciously process the idea that Kyle was having an
       affair


                          ESSLLI-2007 Figurative Language Processing                                  53

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:102
posted:10/2/2011
language:English
pages:9