Docstoc

Slide 1 - 国際開発研究科

Document Sample
Slide 1 - 国際開発研究科 Powered By Docstoc
					JPN494: Japanese Language and Linguistics
JPN543: Advanced Japanese Language and
        Linguistics




                      Semantics & Pragmatics
                      (1)
Semantics & Pragmatics

   Semantics & Pragmatics: the subfields of
    linguistics that deal with meanings.
   Semantics (mainly) deals with “literal
    meaning” = meaning intrinsic to linguistic
    expressions
   Pragmatics deals with a different kind of
    meaning, which results from the interaction
    between linguistic expressions and the
    context in which they are used.
   “Ken doesn’t eat meat.”

   “We are going to a steak house. Should we ask Ken too?”
   “What kind of pizza should we order?”
   “Is Ken health-conscious?”

In different situations, “Ken doesn’t eat meat” has an
invariable literal meaning (explicature) but conveys
different messages (implicature)

What is said vs. What is meant
The boundary between semantics and
pragmatics is not always clear.

   Ken stopped smoking (presupposition)
   KEN is a vegetarian vs. Ken is a
    VEGETARIAN (focus)
   This vase is expensive, that one is not
    (deixis)
Various approaches to linguistic
meaning

   “Meaning” is a vague and abstract notion.

   What is the meaning of “dog”? Is it:
    (i) objects that are called dog?
    (ii) the concept (that people have in their head) of
       “dog-ness”?
   What are the meanings of “A dog barked”,
    “Fido is a dog”, etc.?
   What are the meanings/functions of:
    – は (as in: 山田くんは頭がいい)
    – よ (as in: 山田くんは頭がいいよ)
    – the (as in: Joe broke the base)
    …?

   What are the differences between:
    – John left & John has left
    – It’s Joe that Mary loves & Mary loves JOE
    …?
   “Dog” has various connotations:
    –   denotation: domesticated canine
    –   connotation: diligence, loyalty (positive); flattery,
        humbleness (negative)

    What are the social/historical/cultural backgrounds of such
    connotations?

    警察官 (けいさつかん) vs. お巡りさん (おまわりさん) vs.
    ポリ公 (ぽりこう)
    宿屋 (やどや) vs. 旅館 (りょかん) vs. ホテル
   Even for expressions with “concrete” meanings, the
    exact formulation of their meanings is not a trivial
    matter.
    ⇒ there is a certain consensus, however, as to how
    to precisely formulate the meanings of simple
    expressions
   To obtain deeper understanding of more subtle
    aspects of meanings (e.g. discourse-oriented
    meanings), it is often useful to use/extend the
    analytical tools that are proven to be useful for
    simpler cases.
Foundations of denotational semantics

   Frege’s (1892) proposal: there are two types of
    meanings:
    –   Sinn (= sense): informational content or concept
        associated with an expression
    –   Bedeutung (= reference): object designated by an
        expression
   References (denotations) of objects = worldly objects
    that correspond to expressions
    –   E.g., the reference (denotation) of Ichiro Suzuki is Ichiro
        Suzuki.
    –   [[ Ichiro Suzuki ]] = Ichiro Suzuki
Extensional Semantics

   What are the denotations of names like:
    Ichiro Suzuki, Japan, Apollo 13 … ?
    → (specific) individuals
   What are the denotations of:
    –   common nouns: student, book, dog, …
    –   adjectives: nice, red, American, …
    –   intransitive verbs: sings, barks, snores, …
    → sets of individuals (= properties)
   Fido is a dog.
   Fido is nice.
   Fido barks.

Universe of Discourse: {Chris, David, Evan, Fido, Goliath,
Holly}

[[ dog ]] = {Fido, Goliath, Holly}
[[ nice ]] = {Chris, David, Evan, Fido, Holly}
[[ barks ]] = {Fido}
   What are the denotations of:
    –   relational nouns: sibling, friend, supervisor, …
    –   relational adjectives: fond, jealous, nicer, …
    –   transitive verbs: likes, envies, admires, …
    → sets of pairs of individuals (= (2-place)
    relations)
   Fido is a sibling of Holly.
   Fido is nicer than Holly.
   Fido likes Holly.

Universe of Discourse: {Chris, David, Evan, Fido, Goliath,
Holly}

[[ sibling ]] = {…, <Fido, Goliath>, <Goliath, Fido>, …}
[[ nicer ]] = {…, <Fido, Holly>, <Fido, Goliath>, …}
[[ likes ]] = {<Chris, Chris>, <David, David>, …, <Fido, Holly>,
     …}
   What are the denotations of sentences like:
    “Fido is a dog”, “Fido likes Holly”?
    → truth values (true or false; 1 or 0)

    [[ Fido is a dog ]] = 1
    [[ Fido is nice ]] = 1
    [[ Goliath is nice ]] = 0
[[ Fido is a dog ]] = 1
[[ Fido is nice ]] = 1
[[ Goliath is nice ]] = 0
…

Fido is a dog and Goliath is nice
Fido is nice and Goliath is nice

Fido is a dog or Goliath is nice
Fido is nice or Goliath is nice

Chris saw {an oculist/an eye doctor}
Compositionality

   Principle of Compositionality of Meaning:
    The meaning of the whole is determined by
    (i) the meanings of its parts and (ii) the way
    they are put together.

    Fido is a dog
    Fido likes Holly vs. Holly likes Fido
   mapping from syntactic representations to
    meanings:
    1.   meanings of individual expressions
    2.   semantic rules for complex expressions (VP, S,
         etc.)
[S Fido [VP is [NP a dog]]]

#1.
[[ Fido ]] = Fido
[[ dog ]] = {Fido, Goliath, Holly}

#2.
SR1: [[a N]] = [[N]]

#3.
SR2: [[is NP]] = [[NP]]
[S Fido [VP is [NP a dog]]]

#4.
SR3:
[[ N VP ]] = 1 iff [[N]] ∈ [[VP]]
[[ N VP ]] = 0 otherwise
A possible application

   Customer: “I want a smart Pomeranian that can
    serve as a guard dog.”
   Database
    –   Pomeranian: {Fido, Goliath, Tornado, …}
    –   barks: {Fido, Tornado, …}
    –   smart: {Fido, Holly, …}
   Is “Fido is a Pomeranian” true?
    Is “Fido barks” true?
    Is “Fido is smart” true?
[S Fido [VP likes Holly]]

[[ Fido ]] = Fido
[[ Holly ]] = Holly
[[ likes ]] = {<Chris, Chris>, <Chris, Fido>, <David,
    David>, <David, Goliath>, <Evan, Evan>, <Evan,
    Goliath>, <Fido, Fido>, <Fido, Chris>, <Fido, Holly>,
    <Holly, Holly>, <Goliath, Goliath>}
   The meaning of a 1-place predicate can be represented as a
    set of individuals.
   The meaning of a 2-place predicate can be represented as a
    set of pairs of individuals.
   The meaning of a sentence with a 1-place predicate can be
    “calculated” in terms of the membership-relation between
    individuals and sets.
   The meaning of a sentence with a 2-place predicate is trickier;
    one way to avoid the problem is to use “functions” instead of
    “sets” (keeping the basic idea).
   Semantic analyses in terms of “sets” are more intuitive;
    semantic analyses in terms of “functions” are more flexible and
    technically easier to handle.
Properties/Relations as Functions

   A name denotes an individual
   A sentence denotes a truth value

   A 1-place predicate (e.g., dog, nice) denotes
    a characteristic function of a set of individuals
   Function: a relation between two groups of
    entities (domain & range)

    F(x) = 2x + 1

    F(x) = λx [2x + 1]

    λx [2x + 1] (3) = 6 + 1 = 7
F(x)(y) = λx [λy [2x2 + 3y + 1]]

λx [λy [2x2 + 3y + 1]] (1) = λy [3y + 3]

λx [λy [2x2 + 3y + 1]](1)(2) =
λy [3y + 3] (2) = 9
   A characteristic function of a set: a function that yields value 1 when the
    argument (input) is a member of the set, and yield value 0 otherwise

   A property as a set:
    [[ dog ]] = {Fido, Goliath, Holly}

   A property as a function:
    [[ dog ]] = λx [dog (x)] where:

    λx [dog (x)] (Chris) = 0
    λx [dog (x)] (David) = 0
    λx [dog (x)] (Evan) = 0
    λx [dog (x)] (Fido) = 1
    λx [dog (x)] (Goliath) = 1
    λx [dog (x)] (Holly) = 1
[S Fido [VP is [NP a dog]]]

#1.
[[ Fido ]] = Fido
[[ dog ]] = λx [dog (x)]

#2.
SR1: [[a N]] = [[N]]

#3.
SR2: [[is NP]] = [[NP]]
[S Fido [VP is [NP a dog]]]

#4.
SR3’:
[[ N VP ]] = [[VP]] ([[N]])

#4.
SR3:
[[ N VP ]] = 1 iff [[N]] ∈ [[VP]]
[[ N VP ]] = 0 otherwise
   1-place predicates denote functions from
    individuals to truth values
    –   λx [dog (x)]
   2-place predicates denote functions from
    individuals to functions from individuals to
    truth values.
    –   λy [λx [likes (y)(x)]]
[S Fido [VP likes Holly]]

#1
[[ Fido ]] = Fido
[[ Holly ]] = Holly
[[ likes ]] = λy [λx [likes (y)(x)]]
[S Fido [VP likes Holly]]

#2
SR4:
[[ V NP]] = [[V]] ([[NP]])

#3
SR3’:
[[ N VP ]] = [[VP]] ([[N]])
Semantic Types

   individuals (entities): e
   truth values: t
   functions from individuals to truth values:
    <e,t>
   functions from individuals to functions from
    individuals to truth values: <e,<e,t>>
Quantificational NPs

   Some dog barks
   Every dog barks
   Most dogs bark
   No dogs bark

[[dog]] = {Fido, Goliath, Holly}
[[barks]] = {Fido, Holly}

[[dog]] = λx [dog (x)]
[[barks]] = λx [barks (x)]
   Some dog barks
    –  There is at least one individual that is a dog and that barks.
    in other words:
    – The intersection of the set of dogs and the set of “barkers”
       is not empty.
    in yet other words:
    – There is at least one individual a such that
       λx [dog (x)] (a) = λx [barks (x)] (a) = 1

    A quantificational determiner denotes a relation between two
    properties.
   Every dog barks
    –  All individuals that are dogs bark.
    in other words:
    – The set of dogs is a subset of the set of “barkers”.
    in yet other words:
    – For every individual a such that
       λx [dog (x)] (a) = 1, λx [barks (x)] (a) = 1
   Most dogs bark
    –   .5A > A - B
        where A = the number of dogs & B = the number
        of barkers
   No dogs bark
    –   The intersection of the set of dogs and the set of
        “barkers” is empty
[S [NPSome dog] [VP barks]]

#1
[[ dog ]] = λx [dog (x)]
[[ barks ]] = λx [barks (x)]
[[ some ]] = λP [λQ [SOME(P)(Q)]]

a relation between properties = a function from sets of
individuals to functions from sets of individuals to truth
values

   <<e,t>, <<e,t>, t>>
[S [NPSome dog] [VP barks]]

#2
[[D N]] = [[D]] ([[N]])
= λQ [SOME(λx [dog (x)])(Q)]

#3
[[NP VP]] = [[NP]] ([[V]])
= SOME(λx [dog (x)])(λx [barks (x)]) = 1
   The dog barks
    –   The set of dogs (i) has only one member, and (ii)
        is a subset of the set of barkers.
not, and, or

   Some dog barks and Fido is a dog.
   Fido is a dog or Goliath likes Fido.
    <t, <t, t>>
   It is not the case that Fido is a dog.
    (Fido is not a dog)
    <t, t>
Intensional Semantics

   Limitations of extensional semantics
    –   Sentences can have only two meanings. (All true
        statements cannot be distinguished from each other.)
    –   Expressions that denote the same object cannot be
        distinguished.

        e.g. Even if all dogs are hairy and all hairy creatures are
        dogs, we would not consider “hairy” and “dog” synonyms.
        (cf. oculist - eye doctor, photocopier - Xerox machine)
   Fido will bark.
   Fido barked.

   Chris must be home by now.

   Chris may not come.

   Chris believes that Goliath barked.

   If Goliath was nice, then Chris would like it.
Current vs. Future/Past States of
Affairs

the current state of affairs:

universe: {c, d, e, f, g, h}

[[ student ]] = {c, d, e}
[[ dog ]] = {f, g, h}
[[ nice ]] = {c, d, e, f, h}
…
[[ like ]] = {<c, c>, …, <f, g>, }
…
the state of affairs in one year:

[[ student ]] = {d, e}
[[ dog ]] = {f, g, h}
[[ nice ]] = {c, d, e, f}
…
[[ like ]] = {<c, c>, …}
…
A hypothetical state of affairs

a hypothetical state of affairs:

[[ student ]] = {c, f}
[[ dog ]] = {f, g, h}
[[ nice ]] = {f, g, h}
…
[[ like ]] = {<c, c>, …}
…

the actual state of affairs = the actual world
hypothetical states of affairs = possible worlds
Extension vs. Intension

   Certain expressions make reference to
    future/past times & hypothetical situations.

<w1, t1>      <w2, t1>    <w3, t1>   …
<w1, t2>      <w2, t2>    <w3, t2>   …
<w1, t3>      <w2, t3>    <w3, t3>   …
…                  …           …
   An expression has different denotations in different world/time pairs

at <w1, t1>: [[ student ]] = {c, d, e}
at <w1, t2>: [[ student ]] = {c, d, e}
at <w1, t3>: [[ student ]] = {d, e}
…
at <w2, t1>: [[ student ]] = {c, d, e, f}
at <w2, t2>: [[ student ]] = {c, d, e, f}
at <w2, t3>: [[ student ]] = {d, e, f}
…
at <w3, t1>: [[ student ]] = {c, d}
at <w3, t2>: [[ student ]] = {d, e}
at <w3, t3>: [[ student ]] = {e, f}
…
   An expression has different denotations in different world/time pairs

at <w1, t1>: [[ Chris is a student ]] = 1
at <w1, t2>: [[Chris is a student ]] = 1
at <w1, t3>: [[Chris is a student ]] = 0
…
at <w2, t1>: [[Chris is a student ]] = 1
at <w2, t2>: [[Chris is a student ]] = 1
at <w2, t3>: [[Chris is a student ]] = 0
…
at <w3, t1>: [[Chris is a student ]] = 1
at <w3, t2>: [[Chris is a student ]] = 0
at <w3, t3>: [[Chris is a student ]] = 0
…
   Extension: the denotation of an expression in
    a given situation (a world/time pair).
   Intension: a function from situations to
    extensions
the extension of “student” at <w1, t1> =
[[ student ]]w1,t1 = {c, d, e}

the intension of “student” =
[[ student ]] = the function F such that:

F (<w1, t1>) = {c, d, e}
F (<w1, t2>) = {c, d, e}
F (<w1, t3>) = {d, e}
…
F (<w2, t1>) = {c, d, e, f}
…
   Two expressions may have the same extension in a
    certain situation, without having the same intension.

[[ dog ]]w1,t1 = {f, g, h}
[[ hairy ]]w1,t1 = {f, g, h}

[[ dog ]]w2,t1 = {f, g, h}
[[ hairy ]]w2,t1 = {f, g}
[[ Chris is a student ]]w1,t1 = 1
[[ Fido is nice ]]w1,t1 = 1

[[ Chris is a student ]]w2,t1 = 1
[[ Fido is nice ]]w2,t1 = 0

the intension of a sentence: a function from
situations to truth values

“Chris is a student” is true = “Chris is a student” is true in this
world, at this time
   Sense: concept associated with an expression
   Reference: object designated by an expression

Denotational interpretation of Frege’s sense:
intension = sense
extension = reference

The definition of “sense” in more concrete terms.
   [[ Fido will bark ]]wn,tn = 1 iff there is some time tm
    such that m > n and [[ Fido barks (is barking) ]]wn,tm =
    1; otherwise = 0
   [[ Fido barked ]]wn,tn = 1 iff there is some time tm such
    that m < n and [[ Fido barks (is barking) ]]wn,tm = 1;
    otherwise = 0

<w1, t1> : Fido is not barking
<w1, t2> : Fido is barking
<w1, t3> : Fido is not barking
   [[ Chris must be home (by now) ]]wn,tn
    = 1 iff for most worlds wm that are similar to wn,
    [[ Chris is home (by now) ]]wm,tn = 1;
    = 0, otherwise

   [[Chris may be home ]]wn,tn
    = 1 iff there is some world wm that is similar to wn
    and
    [[ Chris is home ]]wm,tn = 1;
    = 0 otherwise
   [[ If Goliath was nice, then Chris would like
    it ]]wn,tn
    = 1 iff for each world wm such that
    [[ Goliath is nice ]]wm,tn = 1,
    [[ Chris likes Goliath ]]wm,tn = 1
    = 0 otherwise

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:13
posted:7/2/2011
language:English
pages:56