WHAT‘S IN A WORD? - Daniel Wiechmann · WHAT‘S IN A WORD? A INTRODUCTION TO COGNITIVE SEMANTICS Summer Term 2008 –Daniel Wiechmann. Goal of this course Provide you …

WHAT‘S IN A WORD?A INTRODUCTION TO COGNITIVE SEMANTICS

Summer Term 2008 – Daniel Wiechmann

Goal of this course

�Provide you with a cognitively plausible apparatus for the description of linguistic meaningdescription of linguistic meaning

�Connect linguistics to neighboring disciplines

Course prerequisites

� Introductions to Linguistics I & II

� Morphosyntax is only marginally relevant...

� ...but a decent understanding of key concepts used in semantic description is concepts used in semantic description is required

Course requirements

� Weekly reading assignments

� some homework assignments

� oral presentation (20 min)

• term paper• obviously not empirical

• theoretical discussion of some concept discussed here

• meta-analysis (summarize and comment on empirical work that has been done on topic)

• application: use descriptive apparatus introduced here to describe the semantics of a (set of) term(s)

Downloadable Materials

www.daniel-wiechmann.net

↷↷↷↷ FilesLogin : guest

Password : go

Right now you can get:Right now you can get:• Presentation (PPT) for:

Introduction to Linguistics I: Meaning and Use

• Syllabus • Tool: ppt(x) viewer

COURSE PROGRAMME (SYLLABUS)

THE BIG PUZZLE OF MEANINGMEANING

VIEWS FROM COGNITIVE SCIENCE

Studying the human mind is somewhat problematic

Philosophy

Anthropology

Biology

Computer Science

Linguistics

COGNITIVE SCIENCE

Anthropology

Psychology[...]

COGNITIVE SCIENCE

VIEWS FROM PHILOSOPHY

� Meaning and Truth (objectivist approach)� meanings = abstract logical objects

(Frege, Russell, Wittgenstein I, Kripke)

a. There is a real world consisting of entities structured according to their properties and relations. Categorization of these entities is based on their properties.of these entities is based on their properties.

b. The real world is fully and correctly structured so that it can be modeled

c. Symbols are representations of reality and can only be meaningful to the degree that they correspond to reality.

d. The human mind processes the abstract symbols in a computer-like fashion so that it mirrors nature.

e. Human thought is symbol- manipulation and it is independent of the human organism.

http://plato.stanford.edu/

Joint Session of Mind and the Aristotelian Society (1993)

Left to right: W.V. Quine, Donald Davidson, Sir Peter Strawson,Rudolf Fara and Martin Davies

VIEWS FROM AI (ARTIFICIAL INTELLIGENCE)

Neural nets

Knowlegde representation

Scripts

VIEWS FROM COGNITIVE PSYCHOLOGY

Perception,

Categorization

Attention

Gestalts

Figure/ ground

Perception,

Language, and Vision

memory

affordances

Prototype effects

VIEWS FROM COGNITIVE SCIENCE

� Meaning and Cognition� Meaning is grounded in perception and action

� Sensory-motor interaction unites organism with environmentorganism with environment

• connectionism

• dynamic systems theory, non-linear dynamics

• chaos, complexity

• self-organization, emergence

Unifying the picture

Cognitive Grammar

Cognitive Grammar

� Cognitive Linguistics and Cognitive Grammar are “cognitive” in the sense that, insofar as possible, they see language as drawing on other, more basic systems and abilities more basic systems and abilities

� (e.g. perception, attention, categorization)

Cognitive Linguistics

Ronald Langacker:

Cognitive Grammar

Len Talmy:

Force dynamics

Gilles Fauconnier:

Chuck Fillmore & Paul Kay:

George Lakoff:

Metaphor

Idealized cognitive models (ICM)

Lakoff (1987) Women, Fire and Dangerous Things

Cognitive Grammar

Trajector/Landmark organisation

Force dynamics

Figure/Ground organisation

Fauconnier:

Mental Spaces

Conceptual blending

Paul Kay:

Construction Grammar


George Lakoff:

Metaphor Ronald Langacker:

Len Talmy:

Force dynamics

Gilles Fauconnier:

Mental Spaces

Chuck Fillmore & Paul Kay: Metaphor

Idealized cognitive

models (ICM)

Ronald Langacker:

Cognitive Grammar


Langacker (1987) Foundations of Cognitive

Grammar

Force dynamics


Mental Spaces

Conceptual blending

Paul Kay:



George Lakoff:

Metaphor

Ronald Langacker:

Cognitive GrammarGilles Fauconnier:

Mental Spaces

Chuck Fillmore & Paul Kay: Metaphor


Cognitive Grammar


Len Talmy:

Force dynamics


Talmy (2000) Towards a Cognitive Semantics

Mental Spaces

Conceptual blending

Paul Kay:



George Lakoff:

Metaphor

Ronald Langacker:

Cognitive Grammar

Len Talmy:

Force dynamics Gilles Fauconnier:Chuck Fillmore &

Paul Kay: Metaphor


Cognitive Grammar


Force dynamics


Gilles Fauconnier:

Mental Spaces

Conceptual blending

Fauconnier (1985) Mental spaces: Aspects of meaning

construction in natural language.

Paul Kay:



George Lakoff: Ronald Langacker: Len Talmy:Gilles Fauconnier:

George Lakoff:

Metaphor


Ronald Langacker:

Cognitive Grammar


Len Talmy:

Force dynamics


Gilles Fauconnier:

Mental Spaces

Conceptual blending

Chuck Fillmore & Paul Kay:


Frame Semantics


George Lakoff:

Metaphor


Ron Langacker:

Cognitive Grammar

(unified framework)

Len Talmy:

Force-dynamics Figure/Ground organisation

I. Meaning reduces to conceptualization (mental experience)

II. A frequently-used expression typically displays a network of interrelated senses

III. Semantic structures are characterized relative to

Five hypotheses about meaning

III. Semantic structures are characterized relative to cognitive domains

IV. A semantic structure derives its value through the imposition of a profile on a base

V. Semantic structures incoporate conventional imagery, i.e. they construe a situation in a particular fashion

Things we beed to have in place

What is a word?� sense versus reference – Sinn/Bedeutung – intension/extension� referring� predicating

SENSE/LEXICAL RELATIONSSENSE/LEXICAL RELATIONS� Entailment� Opposition� Homonymy� Polysemy� Meronymy� Hyponymy� Ambiguity� Vagueness

Try and keep in mind theses things...

Ludwig Wittgenstein 1889 - 1951

Try and keep in mind these things...

� How is it possible that we are able to talk about reality?� If language were to represent reality, then language and reality must have something in common

� If sentence were to represent states of affairs, then a sentence and the state of affairs that it is about must have something in common

� Wittgenstein (I): correspondence � Wittgenstein (I): correspondence � Structural similarity (structure of sentence mimics structure of fact)

� picturing, mirroring

� Because the world determines the structure of language, we can read off the structure of reality from the structure of language

� There is a cat on the mat

� There is no cat on the mat

[EVANS&GREEN:157 - 1 79 ]

Introducing Cognitive Semantics

THE MEANING OF A WORD CAN BE BROKEN DOWN TO A SET OF MORE BASIC NOTIONS

(FEATURES)

MEANINGS AS FEATURE BUNDLES

(FEATURES)

DECOMPOSITIONAL APPROACH

CLASSICAL APPROACH

Describing concepts

BACHELOR(X)

x is humanx is adultx is male

Necessary and (jointly) sufficient conditions

x is unmarriedx has never been married

Describing concepts:Meaning Postulates

BACHELOR

For all x, x is a woman if and only if: L

Where L is a list of features|attributes|properties|conditions

L = { [HUMAN] & [ADULT] & [MALE] L = { [HUMAN] & [ADULT] & [MALE] & ¬ [MARRIED] }


BACHELOR



L = { [HUMAN] & [ADULT] & [MALE] & ¬ [MARRIED] }

(i) John is a bachelor (i) John is a bachelor entails

(ii) John is a man

(i) John is a bachelorentails and is entailed by (i.e. is synonymous with)

(iii) John is a man who has never married

(i) John is a bachelorcontradicts

(iv) John is married


BACHELOR



L = { [HUMAN] & [ADULT] & [MALE] & ¬ [MARRIED] }

intersection inclusion

Problems with necessity

BACHELOR(X)

x is HUMANx is ADULTx is male

x is unmarriedx is unmarriedx has never been married

some concepts are ill-defined:the case of game

LINK TOWITTGENSTEIN (II) PU

Encyclopedic knowledge & extended meanings

i. The child is safeii. The beach is safeiii. The shovel is safe

i. Watch out Jane, your husband‘s a right bachelor

Words (expressions), Sense, and Meanings

Polysemy- one form, many meanings -- one form, many meanings -

Polysemy: An examplea ring (=N)

CIRCULARPIECE OF

JEWELRY FOR

FINGER

basic sense


RECTANGULAR

ARENA

CIRCULAR

ARENA

CIRCULAR

ENTITY

GROUP OF PEOPLE

OPERATION TOGETHER

(SECRETLY)

ARENA

CIRCULARPIECE OF

JEWELRY FOR

FINGER

CIRCULAROBJECT

CIRCULARMARK

CIRCULARPIECE OF

JEWELRY

= x is schematic of y

basic sense

= x instantiates y

= x is more peripheral than y

= x is more central than yCENTRALITY/PERIPHERALITY

GENERALITY/SPECIFICITY


RECTANGULAR

ARENA

CIRCULAR

ARENA

CIRCULAR

ENTITY

GROUP OF PEOPLE

OPERATION TOGETHER

(SECRETLY)

e.g.

CIRCULAROBJECT

CIRCULAR MARK

e.g.

METAPHORICAL

EXTENSION(senses are linked via similarity)

or:

METONYMICAL

EXTENSION(senses are linked via association; church)

Frames, Domains, Schemas, Scripts, ...

Frames:

•Notion from Frame Semantics (Fillmore 1985 )

•Fillmore views his frame semantic model as a model of the semantics of

understanding

•In F view, speakers produce words and constructions in a text as tool for a

particular activity, namely to evoke understanding (=tool metaphor of

understanding process)understanding process)

•Hearer‘s task is to figure out the activity these tools were intended for

•Words evoke (activate) frames , i.e. knowledge structures about types of

situation

•Sell evokes COMMERCIAL TRANSACTION frame

•one cannot understand the meaning of sell without access to the

knowledge about commercial transactions, which involves particular

entities that stand in particular relations to each other (buyers, goods,

money,....)

•Frames are based on recurrent experience


Scripts:

•Notion from Script theory in artificial intelligence (Schank 1982, 1986)

•Schank attempted to model knowledge representation (especially in the

context of language understanding)

•All memory is episodic, i.e. organized around a person‘s experience

•Generalized episodes are called scripts

•Words evoke (activate) scripts, i.e. knowledge structures about types of •Words evoke (activate) scripts, i.e. knowledge structures about types of

situation

•Scripts allow individual to make inferences

•Scripts allow individual to make predictions

•Scripts allow individuals to fill in knowledge (cf. pragmatic enrichment)


•Words evoke (activate) scripts, i.e. knowledge structures about types of

situation

•Example: Restaurant script

•Scene 1: entering

•S PTRANS S into RESTAURANT

•S ATTEND eyes to TABLES

•S MBUILD where to sit•S MBUILD where to sit

•S PTRANS S to TABLE

•S MOVE to SITTING position

•Scene 2: Ordering

•S PTRANS MENU to S

•S MBUILD choice of FOOD

•S MTRANS signal to WAITER

•WAITER PTRANS WAITER to TABLE

•S MTRANS P(p=I want food) to WAITER

•WAITER PTRANS WAITER to COOK

•Scene 3: Eating [...]

•Scene 4: Exiting [...]

Langacker: content, construal

� Meaning of an expression E consists of conceptual

content (CONTENTE) and a particular way of construing

(viewing) that content (CONSTRUALE)

An expresssion‘s content is provided by a set of cognitive � An expresssion‘s content is provided by a set of cognitive

domains (complex matrix), in each of which the

designated entity plays some role

Profiling

Within the array of conceptual content it evokes as the

basis for its meaning (its conceptual base), an expression

profiles (i.e. refers to) a particular substructure.

hypotenuse

Profiling

Within the array of conceptual content it evokes as the

basis for its meaning (its conceptual base), an expression

profiles (i.e. refers to) a particular substructure.

auntEGO

+FEMALE

Profiling

Expressions that evoke the same content may contrast

semantically by virtue of their choice of profile within

this base

M F

base

Profiling



this base

M F

base

wife

Profiling



this base

M F

base

husband

First thoughts about: image schemas & conceptual metaphors

VISUAL PERCEPTION

BASIC PROCESSES

Auditory and Visual Pathways and Levels of Abstraction

Visual system

Visual pathways

Visual processing

Perceptual organisationGestaltist approaches

� Group of German psychologist� Koffka, Köhler, Wertheimer

� Fundmental principle of perceptual organisation:

Koffka

� Fundmental principle of perceptual organisation:

� Law of Prägnanz

� Of several geometrically possible organisations that one will actually occur which possesses the best, simplest, and most stable shape (Kurt Koffka 1935)

Gestalt laws – organisational principles

Figure ground organisation

Top-down processing

Hierarchical organisation of human figure(Marr & Nishahara 1978)

Recognition by componentsBiedermann‘s Geons

Pandemonium model

Marr‘s computational theory

INTRODUCING

What do meanings look like?

INTRODUCING IMAGE SCHEMAS

Embodiment and conceptual structure

EMBODIMENT

CONCEPTUAL STRUCTURE

SEMANTIC STRUCTURE

What does meaning look like?

� IMAGE SCHEMA

� schematic and imagistic concepts which

� are abstracted from pre-conceptual bodily experience

� function as constituents of more complex notions

provide the structure projected metaphorically to more abstract � provide the structure projected metaphorically to more abstract domains

Image schema

� Image -> imagistic (adj.) ~ sensory experience� Sensory systems: visual, auditory, haptic, tactile, vestibular system

� Contrasts with introspective experience (feelings, emotions)

� Schema -> schematic (adj.) ~ not detailed, but abstractSchema -> schematic (adj.) ~ not detailed, but abstract� Result from repetead experienc e of similar situation

� Example: CONTAINER (cf. Example w/ in, out)

� Image schemas are abstracted from experience and, once established, serve a categorizors (AS-IF description, -> conceptual metaphors)


� She`s in love

� We‘re out of trouble now

� She‘s coming out of a coma� She‘s coming out of a coma

� I‘m slowly getting into shape

� She entered a state of euphoria

� She fell into a depression


� She`s in love� EMOTIONAL STATES are CONTAINERS

� We‘re out of trouble now� SITUATIONS are CONTAINERS

� She‘s coming out of a coma� She‘s coming out of a coma� BODILY STATES are CONTAINERS

� I‘m slowly getting into shape� BODILY STATES are CONTAINERS

� She entered a state of euphoria� EMOTIONAL STATES are CONATINERS

� She fell into a depression� EMOTIONAL STATES are CONTAINERS

IMAGE SCHEMAS, CONCEPTUAL STRUCTURE and SEMANTIC STRUCTURE

� Image schemas give rise to types of objects (container)/dimensions of contrast (up/down)

� they can also be more complex and structured entities (SOURCE-PATH-GOAL schema)

� can be viewed as the cognitive foundation of FRAMES, � can be viewed as the cognitive foundation of FRAMES, SCRIPTS

� function as categorizors (event may be construed as instantiating a particular image schema)

IMAGE SCHEMAS (examples)

� Container� Blockage� Enablement� Source-Path-Goal� CyclePart-whole

� Attraction� Link� Near-far� Merging� Matching� Contact

� Part-whole� Full-empty� Iteration� Surface� Balance� Counterforce� Process� Collection

� Contact� Object� Compulsion� Restraint removal� Mass-count� Center-periphery� Scale� Splitting� Superimposition

� Container� Blockage� Enablement� Source-Path-Goal� Cycle� Part-whole� Full-empty� Iteration� Surface� Balance� Counterforce

Process

IMAGE SCHEMAS (examples)

� Process� Collection� Attraction� Link� Near-far� Merging� Matching� Contact� Object� Compulsion� Restraint removal� Mass-count� Center-periphery� Scale� Splitting� Superimposition

ANG:FC:277:Joh::1993

Term paper topics:Critical evaluation of theoretical constructs

� A closer look at prototype theory (PT)� Problem of prototypical primes

� Problem of ignorance and error

� Missing prototypes problem

� Problem of compositionality

� Start with general description of PT

� Present challenges of PT

� Focus on subset of problems

� Have a closer look at the texts that formulated the criticism (original sources, not textbook summaries)

� Evaluate the criticim (discuss the problems)

ACCOUNTING FOR TYPICALITY EFFECTS

Idealised cognitive models

ICM –what are hey

� ICM (Lakoff 1987)

� relatively stable mental representations that represent theories about the world

� compare THEORY-theory of concepts ( Margolis & Laurence 1999, Concepts: Core readings )

� KNOWLEDGE APPROACH to concepts ( Murphy 2004‚ The big book of concepts )

� Similar to Fillmore‘s FRAME s(both relate to complex world knowledge)

� Idealised ~ generalized (i.e. abstractions over instances)

ICM – What do they do?

� ICM guide cognitive processes

� Categorization of perceived stimulus

� ad hoc categories (e.g. THINGS TO DO WHEN YOUR LEG FALLS ASLEEP) exhibit prototype effects; theory of ICM aims at accounting for that

� Reasoning (=draw inferences)� Reasoning (=draw inferences)

ICM – How do they give rise to prototype effects?

� Typically effects arise due to mismatches between ICMs against which particular concept is understood� goodness-of-example rating experiment

� Method: ask people to representativeness of item x for category y

� Empirical result: item pope is judged atypical for category BACHELOR (low scores)

� ICM explanation for that finding:

� BACHELOR is understood against MARRIAGE frame/ICM

� the concept POPE is understood (primarily) against CATHOLIC CHURCH frame/ICM� → Clash of cognitive models (=background knowledge about MARRIAGE & CATHOLIC CHURCH) which is used to categorise perceived objects

� → POPE is mentally categorised as peripheral member of category BACHELOR

� → POPE is judged to be a poor example in experiment (=typicality effect)

THOUGHT IS STRUCTURED NEURAL ACTIVITY

Some conceptual background on how the body computes

NEURAL ACTIVITY

LANGUAGE IS INEXTRICABLE FROM THOUGHT AND EXPERIENCE

(SUBSTANCE) DUALISM –> BODY AND SOUL

Some common (mis)conceptions

BODY AND SOUL

BRAINS AND MIND

PHYSICAL STATES AND MENTAL STATES

CENTRAL METAPHOR:

WE ARE LIKE COMPUTERSBRAINS ~ HARDWARE

MINDS ~ SOFTWARE

Some quick facts about neural information processing

� BRAINS

� 100,000,000,000 processing units

� 1000 operations per second

embodied

� DIGITAL COMPUTER

� 1-100 processing units

� 1,000,000,000 operations per second

� abstract, disembodied

frequently crashes� embodied

� fault tolerant

� gradient, probabilistic signals

� evolves and is self organizing

� learns

� frequently crashes

� binary, deterministic signals

� is explicitly designed

� is programmed

How the body computes...in a nutshell


carries out vital life functions

function as INPUT TERMINALS

functions as single OUTPUT CABLE




Artificial neural nets

Interactive activation model

(Partial) Semantic Network (Quillian 1969)

STOP HERE AT THE STOP HERE AT THE LATEST

WHAT‘S IN A WORD? - Daniel Wiechmann · WHAT‘S IN A WORD? A INTRODUCTION TO COGNITIVE SEMANTICS Summer Term 2008 –Daniel Wiechmann. Goal of this course Provide you …

Documents