Lexical-Functional Grammar - University of Cambridgews390/course/syntax/lfgarch.pdf · 2020. 2. 5. · Lexical-Functional Grammar The idea that words and phrases are alternative means

Lexical-Functional GrammarArchitecture

Weiwei Sun

Institute of Computer Science and TechnologyPeking University

September 15, 2018

Grammatical Functions Syntactic Descriptions Structural Correspondences Configurational vs. Nonconfigurational

Homework: 读读读书书书报报报告告告

挑一篇阅读材料，写一篇内容总结（中文），要求：

I 在理解的基础上概述阅读材料的最主要内容

I 不超过2000字（不包括图、公式等）

I 3月26日前电邮PDF版发送给我

Reference

I Chapter 1 of Lexical-Functional Syntax: Motivation for theLFG Architecture.

I Ash Asudeh, Mary Dalrympl, and Ida Toivonen.Constructions with lexical integrity. Journal of LanguageModelling.

I Mary Dalrymple, Ronald M. Kaplan, and Tracy HollowayKing. Economy of Expression as a principle of syntax.Journal of Language Modelling.

Weiwei Sun Lexical-Functional Grammar 2/65


Word groups or word shapes (1)

S

VP

VP

NP

that dog

V

chasing

Aux

are

NP

the two small children

(1) a. *The two small are chasing that children dog.

b. *The two small are dog chasing children that.

c. *Chasing are the two small that dog children.

d. *That are children chasing the two small dog.



Word groups or word shapes (2)

S

NP

malikidog

.ABS

NP

kurdu-jarra-rluchild

-DUAL-ERG

NP

yalumputhat.ABS

V

wajili-pi-nyichase

-NPAST

Aux

ka-palapres

-3DU.SUBJ

NP

wita-jarra-rlusmall

-DUAL-ERG

Warlpiri

Every permutation of the words in the sentence is possible, solong as the auxiliary tense marker occurs in the second position.

Bresnan (1998, 2001)

Morphology competes with syntax.

Absolutive: Morphologicl case in ergative languages for indicating subject of

intransitive verbs and object of transitive verbs. Ergative: Morphologicl case in

ergative languages for indicating agent of the transitive verbs in the basic voice.Weiwei Sun Lexical-Functional Grammar 4/65


Grammatical relations

Although Warlpiri lacks English-style phrase structure, and En-glish lacks Warlpiri-style case and agreement forms of words, theyhave a common organization at a deeper level.

Binding: Subject/Object matters

(2) a. Lucy is hitting herself.

b. *Herself is hitting Lucy.

(3) a. Napaljarri-rli ka-nyanu paka-rni.Napaljarri-ERG PRES-REFL hit-NONPAST.‘Napaljarri is hitting herself.’

b. *Napaljarri ka-nyanu paka-rni.Napaljarri.ABS PRES-REFL hit-NONPAST.‘Herself is hitting Napaljarri.’



Discussion

Do you like the following representation?

S

Predicate

VP

Object

NP

N

boy

Det

the

Main Verb

V

frighten

Aux

M

may

Subject

NP

N

sincerity



Discussion

Do you like the following representation?

S

Predicate

VP

Object

NP

N

boy

Det

the

Main Verb

V

frighten

Aux

M

may

Subject

NP

N

sincerity



F-structure: motivation

Assumption

For any language functional syntactic concepts such as subjectand object are relevant.

⇒Use f(unctional)-structure

I to represent what languages have in common inwide-spread phenomena,

I to capture some universal properties of language

There is no advantage in representing such information as phrase-structure information.



Discussion

How should abstract grammatical relations be captured?

I Transformational Grammar: configurationally, using auniform syntactic representation

I LFG: nonconfigurationally, using a separate syntacticrepresentation

What does this representation mean?

(4) sincerity may frighten the boy.

pred ‘我⟨subj,obj

⟩’

subj[pred ’sincerity’

]

obj

pred ’boy’

def +

pers 3rd

num sg



Discussion

How should abstract grammatical relations be captured?

I Transformational Grammar: configurationally, using auniform syntactic representation

I LFG: nonconfigurationally, using a separate syntacticrepresentation

What does this representation mean?

(5) sincerity may frighten the boy.

pred ‘我⟨subj,obj

⟩’

subj[pred ’sincerity’

]

obj

pred ’boy’

def +

pers 3rd

num sg



Lexical-Functional Grammar

The idea that words and phrases are alternative means of ex-pressing the same grammatical relations underlies the design ofLexical-Functional Grammar (LFG).

Pioneers

Developed in the late 70s by Joan Bresnan and Ron Kaplan

I J. Bresnan: A linguist trained at the MIT

I R. Kaplan: A psycholinguist and computational linguisttrained at Harvard



Demo

http://clarino.uib.no/iness/xle-web


http://clarino.uib.no/iness/xle-web


Outline

Grammatical Functions

Syntactic Descriptions

Structural Correspondences

Configurational vs. Nonconfigurational



Grammatical functions

Universally-available grammatical functions

LFG posits a universal inventory of grammatical functions.

I subj: subject

I obj: object

I comp: sentential or closed (nonpredicative) infinitivalcomplement.

I xcomp: an open (predicative) complement, ofteninfinitival, whose subj function is externally controlled.

I objθ: a family of secondary objfunctions associated witha particular, language-specific set of thematic roles

I oblθ: a family of thematically restricted oblique functions.E.g. oblGOAL, oblSOURCE,

I adj, xadj: adjunct functions



Subject (1)

Case

(6) he/*him broke the window

Agreement

(7) I am / You are / He is

Subjecthood test

How about Mandarin Chinese?



Object

I Many languages have more than one phrase bearing an objectfunction.

(8) a. He gave her a book.b. I bet you 1 million pounds you won’t click on this

video. (from YouTube)

I Languages allow a single thematically unrestricted object, theprimary obj.

(9) a. I gave her a book.b. I gave a book to her.

I Languages may allow one or more secondary, thematicallyrestricted objects, viz objθ. In English, the thematicallyrestricted object must be a theme.

(10) a. I made her a cake.b. *I made a cake her.



Oblique

I Obliques are with an explicit indication of the thematic role.In English, this indication is by means of prepositions: obliquearguments are PPs, while objects are bare NPs/DPs.

I The oblique argument functions include such grammaticalfunctions as GOAL (to), BENEFACTIVE (for), SOURCE(from), INSTRUMENT (with), LOCATION (variousprepositions), AGENT (by), etc.

I oblθ: oblgoal, oblben, oblsource, oblinstrument, oblloc,oblagent

(11) a. David gave the book [oblgoal to Chris].

b. *David gave the book to Chris to Ken.



COMP/XCOMP/XADJ

The comp, xcomp, and xadj functions are clausal functions.

I The comp function is a closed function containing an internalsubj phrase.

(12) a. David complained that Chris yawned.b. David wondered who yawned.c. David couldn’t believe how big the house was.

I The xcomp and xadj functions are open functions that donot contain an internal subject phrase. Their subjmust bespecified externally to their phrase.

(13) a. David seemed to yawn.b. Chris expected David to yawn.

(14) Stretching his arms, David yawned.



Cross-classification of grammatical functions

Several cross-classifications are possible among grammaticalfunctions.

I Governable functions: subj, obj, (x)comp, objθ, oblθare governed or subcategorized for by the predicate.

I Modifiers: adj, xadj modify the phrase they appear in,but they are not subcategorized for by the predicate.

I Open functions: xcomp, xadj

I Closed functions: subj, obj, comp, objθ, oblθ, adj



Classification of the argument functions

Core vs. Non-core

I Core arguments/terms: subj, obj, objθI More strictly grammatical functions

I Non-core functions/non-terms: oblθI Oblique elements are much less active syntactically than

core elements.

Restricted vs. Unrestricted

I Semantically unrestricted functions: subj, obj

I Semantically restricted functions: objθ, oblθ



Nongovernable grammatical functions

There are two nongovernable grammatical functions:

I adj: grammatical function of modifiers

I xadj: open predicative adjuncts, whose subj function isexternally controlled.

(15) a. Having opened the window, David took a deepbreath.

b. David ate the celery naked.c. David ate the celery raw.

More than one adjunct function can appear in a sentence.

(16) David devoured a sandwich at noon yesterday.



The autonomy of functional organization

Key: GFs are primitive concepts

LFG does not assume that abstract grammatical functions aredefined in terms of their phrase structural position in the sentencenor in terms of morphological properties like casemarking.

Comparison to GB

Grammatical relations are defined structurally, in terms of thetree.

I Subject: NP or CP daughter of TP

* Object: NP or CP daughter of a VP headed by atransitive verb

* Object of preposition: NP daughter of PP



Examples of English grammatical functions in LFG(based on Asudeh and Toivonen 2015)

subject Some people with no shame walked in and wreckedthe party.

(subj) The party was wrecked by some people with noshame.

object Primary object(obj) Ricky trashed the hotel room.

Ricky gave John a glass.Ricky gave a glass to John.

objectθ Secondary object; thematically restricted object(object theme, restricted to theme roles)(objθ in English restricted to theme, cannot be ben-eficiary)

(objθ) Sandy gave John a glass.Tom baked Susan a cake.




obliqueθ Typically has oblique case or is a PP(oblθ) Julia placed the vase on the desk.

Ricky gave a glass to John.

complement Closed (saturated) complement: a clausal argu-ment which has its own subject

(comp) Peggy told Matt that she had won the prize.

xcomp Open (unsaturated) predicate complementI told Patrick to quit.Peggy-Sue seems to be a complete fraud.

adjunct A modifier, a nonargument(adj) Mary read a good book.

Mary counted the cars very quickly.Sally killed a bug in the yard.Since she had no money, Mary was forced toget a job.




xadj Open predicate adjunctHaving no money, Mary was forced to get a job.

possessor Possessor phrase(poss) John’s book

topic Grammaticalized discourse function;(top) must be identified with or anaphorically linked

to another grammatical functionBagels, Mary loves. (top = obj)As for bagels, Mary loves them. (top anaphor-ically linked to obj)

focus Grammaticalized discourse function;(foc) must be identified with or anaphorically linked

to another grammatical functionWhich author do the critics praise? (foc =obj)



Outline







Multiple parallel structures

Assumption: language is made up of multiple dimensions ofstructure.

I LFG describes and models language by parallel structures.

I LFG also illustrates how different aspects of linguisticstructure are related.

Traditional LFG analyses focus on two structures

Constituent/categorial structure (c-structure)

I overt, more concrete level of linear and hierarchicalorganization of words into phrases.

Functional structure (f-structure)

I abstract functional syntactic organization of the sentence

I syntactic predicate-argument structure and functionalrelations like subject and object.



C-structure

Described by conventional PS trees

PS trees are defined in terms of syntactic categories, terminalnodes, dominance and precedence.

I They don’t contain unpronounced words.

I They reflect the structure and grouping of words andphrases in the sentence.

Not universal

I Languages are very different on the c-structure level.

I The inventory of phrasal categories is not universally fixed,but may vary from language to language.



Phrase structure rules

CFG rules

S → NP VP

I This rule permits a node labeled S to dominate two nodes,an NP and a VP, with the NP preceding the VP.

I In LFG, phrase structure rules are interpreted as nodeadmissibility conditions: a phrase structure tree isadmitted by a set of phrase structure rules.

LFG employ more flexible phrase structure rules

I IP → {NP|PP} I′

I Either a NP or a PP can appear in the specifier position.I {...|...} mark disjunction.

I VP → V (NP) PP∗

I (NP): an optional NPI ∗: can repeat zero or many times.



ID/LP rules

I Immediate Dominance (ID) rules: dominance relations

I Linear Precedence (LP) rules: precedence constraints

ID rule: Using commas

VP→V, NP

⇒ VP→V NP

⇒ VP→NP V

LP ordering constraint: Using ≺VP→V, NP V≺NP

⇒ VP→V NP

VP→V, NP, PP V≺NP, V≺PP

⇒ {VP→V NP PP | VP→V PP NP}



F-structure representation

Described by functions

I In LFG, functional information is formally represented bythe f-structure.

I Mathematically, the f-structure can be thought of as afunction from attributes to values,

I or equivalently as a set of 〈attribute, value〉 pairs.

Attribute-Value Matrix (AVM)

We can represent f-structures in tabular form[attribute1 value1attribute2 value2

]



An example

Example

(17) I saw the girl.

subj

pred ’pro’

pers 1st

num sg

tense past

pred ’see⟨subj, obj

⟩’

obj

pred ’girl’

def +

pers 3rd

num sg

LFG is functional!



Formal properties of f-structures

Definition (F-structure)

An f-structure is a finite set of pairs of attributes and values.An f-structures attributes can be

I atomic symbols, e.g. subj, obj, pred

An f-structures values can be:

I atomic symbols, e.g. sg , 1st, +, past

I semantic forms, e.g. ’girl’, ’see〈subj, obj〉’I f-structures

I a set of f-structures

Attributes with the same valuesatt1 1

[a1 v1

a2 v2

]att2 1



Example

Example: David yawned quietly yesterday.

subj[pred ’david’

]pred ’yawn

⟨subj

⟩’

tense past

adj

[pred ’quietly’

][pred ’yesterday’

]



Example



]pred ’yawn

⟨subj

⟩’

tense past

adj

[pred ’quietly’


]

Atomic



Example



]pred ’yawn

⟨subj

⟩’

tense past

adj

[pred ’quietly’


]

Semantic form



Example



]pred ’yawn

⟨subj

⟩’

tense past

adj

[pred ’quietly’


]

F-structure



Example



]pred ’yawn

⟨subj

⟩’

tense past

adj

[pred ’quietly’


]

A set of f-structures



Subcategorization

I A semantic form may contain an argument list, next to itssemantic predicate name, e.g.

I ’yawn〈subj〉’I ’see〈subj, obj〉’I ’give〈subj, obj, objtheme〉’

I Lexical items select for grammatical functions (not for NPs,CP, etc)

I How to make sure that subcategorization requirements arefulfilled?

Well-formedness constraints on the f-structure

1. Completeness

2. Coherence

3. Consistency (uniqueness)



Principle of completeness

All governable functions present in the argument list of a seman-tic form must be present in the f-structure.

(18) *He devoured.subj[pred ’pro’

]pred ’devour

⟨subj, obj

⟩’

Definition

Local Completeness: An f-structure is locally complete iff itcontains all the governable functions that its predicate governs.Completeness: An f-structure is complete iff it is locally com-plete and all its subsidiary f-structures are locally complete.



Principle of coherence

All governable functions present in the f-structure are also presentin the argument list of the predicate.

(19) *David yawned the flower.subj

[pred ’David’

]obj

[pred ’David’

]pred ’yawn

⟨subj

⟩’

Definition

Local Coherence: An f-structure is locally coherent iff all thegovernable functions it contains are governed by its predicate.Coherence: An f-structure is coherent iff it is locally coherentand all its subsidiary f-structures are locally coherent.



The principle of consistency (uniqueness)

An attribute has a unique value.

Definition

Consistency: An f-structure is consistent iff all attributes haveat most one value (which may be a set).

The value of the (x)adj function is a set of f-structures:

pred ’devour⟨subj, obj

⟩’


]obj

[spec A

pred ’sandwich’

]

adj

[pred ’quietly’

],

pred ’at⟨obj

⟩’

obj[pred ’noon’

]



Functional description

Function

The language of functional descriptions is based on the mathe-matical conception of f-structures as functions:

f(x) = x2 ⇒ f(SUBJ)= g ⇒ (f SUBJ)= g

Functional description

(f FEAT): the value of the FEAT feature in f

I (f tense) = past

I (f subj) = g

I h ∈ (f adj)



How to construct f-structures?

Functional descriptions just present partial constraints. We findsound f-structures by solving a bunch of constraints, like solvingan algebra problem.(

2 11 2

)(xy

)=

(33

)=⇒

(xy

)=

(11

)

Minimality

The f-structure for an utterance is the minimal solution satisfyingthe constraints introduced by the words and phrase structure ofthe utterance.

Subsumption

A structure A subsumes a structure B iff A and B are identicalor B contains A and additional information not included in A.



An example

Example

(f subj) = g(f pred) = ‘laugh〈subj〉’(g pred) = ‘david’

⇓

f:

pred ‘laugh⟨subj

⟩’

subj g:[pred ‘david’

]

We could enhance our language for functional descriptions.



Negation

Negation

An f-description can be negated; when this happens, the f-description must not be satisfied.

Example

(20) a. I know whether/if David yawned.

b. You have to justify whether/*if your journey is reallynecessary.

justify (f comp compform) 6= IF



Existential constraints

Existential constraint

An f-structure may be required to contain an attribute, but itsvalue may be unconstrained.

Example

(21) a. the man who yawned

b. the man who yawns

c. the man who will yawn

d. *the man who yawning

We can enforce the requirement for relative clauses to be tensedby means of a constraint like the following:(f tense)



Constraining equation

Constraining equation

I Defining equations determine the minimal solution

I Constraining equations check that the minimal solution iswellformed.

Example

(22) a. Chris thought that David yawned.

b. Chris thought David yawned.

c. That David yawned surprised Chris.

d. *David yawned surprised Chris.

I (f compform) = THAT

I (f compform) =c THAT



Inside-out constraint

Outside-in vs. Inside-out

I Outside-in functional constraint:I (f pred) = ‘rock′

I (f case) = LOC

I Inside-out functional constraints:I ((oblloc g) case) = ERGI (subj oblloc g)

Example

(oblloc g) is labeled f : f = (oblloc g)subj f

case ERG

oblloc g

[pred ’rock’

case LOC

]



Questions

1. Where can I find these functional descriptions?

2. What is the relation between c-structure and f-structure?



Outline







Structural correspondences (1)

Question

I C-structures and f-structures represent different propertiesof an utterance.

I How can these structures be associated properly to aparticular sentence?

Codescription

Simultaneously describing more than one structure: Each pieceof the c-structure is directly associated with a description of partof the f-structure.

Example

Sf1

VPf3NPf2f1, f3

[subj f2

]




Correspondence function

A function φ maps c-structures to f-structures φ : N 7→ F .

I φ(n): f-structure associated with n

I φ(M(n)): f-structure associated with the parent node of n

Example

S

VP

V

smiled

NP

N

David

fS, fVP, fV:

pred ’smile⟨subj

⟩’

tense past

subj gNP, gN:

pred ’david’

num sg

pers 3rd




The head convention

A constituent structure phrase and its head map to the samef-structure.

Example

I S, VP and V map to the same f-structure f .

I NP and N map to the same f-structure g.

S

VP

V

smiled

NP

N

David

f :


⟩’

tense past

subj g:

pred ’david’

num sg

pers 3rd



Annotating PS-rules: heads

Consider the rewrite rule: VP → V

Notation

VP and V have the same f-structure by annotating the V-node:

VP → Vφ(M(n)) = φ(n)

The equation indicates that the f-structure of the parent node ofV (i.e. φ(M(n))) is equal to one of the node V (φ(n)).

An alternative notation

I ↑≡ φ(M(n))

I ↓≡ φ(n)VP → V

↑=↓



Annotating PS-rules: grammatical functions

Notation

S → NP VP(φ(M(n)) subj) = φ(n) φ(M(n)) = φ(n)

The first equation indicates that the subj feature of the f-structure of the parent node of NP (i.e. (φ(M(n)) subj)) isequal to the f-structure of NP (φ(n)).


I ↑≡ φ(M(n))

I ↓≡ φ(n)S → NP VP

(↑ subj)=↓ ↑=↓

Key idea

Local description of partial structures



Instantiation

S

VP

V

smiled

NP

N

David

fS, fVP, fV:


⟩’

tense past

subj gNP, gN:

pred ’david’

num sg

pers 3rd

I (fS subj) = fNP)

I fNP = fNI (fN pred) = ‘david’

I (fN num) = sg

I (fN pers) = 3rd

I fS = fVPI fVP = fVI (fV pred) = ‘smile〈subj〉’I (fV tense) = past



Adjuncts

I The attribute adj takes a set as its value

I The c-structure/f-structure correspondance rule expressesmembership to a set.

N → ADJP N↓∈ (↑ adj) ↑=↓

Example

N

N↑=↓

girl

A↓∈ (↑ adj)

pretty

pred ’girl’

num sg

pers 3rd

adj

{[pred ’pretty’

]}



Lexical entries

Notation

In lexical entries, information about the item’s f-structure is rep-resented in the same way:

smiled V (↑ pred) = ‘smile〈subj〉’(↑ tense) = past


The equivalent phrase structure rule:

V → smiled(↑ pred) = ‘smile〈subj〉’

(↑ tense) = past



Lexical integrity

Principle lexical integrity

Morphologically complete words are leaves of the c-structure tree,and each leaf corresponds to one and only one c-structure node.

I Fully inflected words are the terminal elements of thec-structures.

I Every word belongs to exactly one node.

I The structural formation of words is independent of thestructural formation of phrases.

Warlpiri

I The relative order of words in sentences is extremely free.

I The relative order of stems and inflections in words is fixed.



Lexical integrity: A seemly irrelevant illustration

Chomsky (1970) demonstrated that

NPs based on “derived” nouns (i.e. nouns that have verbalcounterparts) have exactly the syntax of NPs based onunderived nouns.

Chomsky reasoned,

we need to recognize that both types are base-generated asnouns instead of attempting to derive certain NPs fromclausal counterparts

Target syntax argument

To generate A directly instead of deriving it from C if there existsa pattern B that has the same target syntax as A and is clearlynot derived from C.

C → AB






Chomsky reasoned,




C → AB






Chomsky reasoned,




C → AB






Chomsky reasoned,




C 9 AB



Inflected words (1)

According to the principle of lexical integrity

I Internal structural formation of words invisible toc-structure principles.

I But the f-structures specified by words is allowed to unifywith the f-structures of the syntactic contexts.

Example: Descriptions

lion (↑ pred) = ‘lion’-s (↑ num) = pl

Example: Structures

N

lions

fN

[pred ‘lion’

num pl

]



Inflected words (2)

Example: Descriptions

live (↑ pred) = ‘live〈...〉’-s (↑ tense) = pres

(↑ subj) =↓(↓ pers) = 3(↓ num) = sg

Example: Structures

VfN

live sfs

fV

pred ‘live

⟨...⟩’

tense pres

subj fs

[num sg

pers 3

]



Example I: Basis

Example

(23) David yawned

P122. Lexical-Functional GrammarOn whiteboard



Outline







Two types of encoding

Grammatical functions are encoded in different ways in differentlanguages. Two fundamental ways for language to realizeunderlying concepts:

I Phrase structure (groups)

I Morphology (shapes)

Morphology competes with syntax

I English: phrase structure strategy (configurational)

I Warlpiri: morphological strategy (nonconfigurational)

Languages may tend to employ one type of encoding more heavily,but there are many cases in which a single language employs mixedstrategies.



English

I subj, obj, and oblθ are primarily encodedconfigurationally.

⇒ PS rules contain specifications of particular grammaticalfunctions.

PS rules

IIP → NP I′

(↑ subj) =↓ ↑=↓

II′ → I VP

↑=↓ ↑=↓

IVP → V′

↑=↓

IVP → V NP

↑=↓ (↑ obj) =↓



Warlpiri (1)

Grammatical function is determined by morphological casemark-ing on the argument phrase

GF≡{subj|obj|objθ|oblθ|comp|xcomp|adj|xadj}

PS rules

I

IP → NP I′

(↑ focus) =↓ ↑=↓(↑ gf) =↓

II′ → I S

↑=↓ ↑=↓

IS → { NP | V }∗

(↑ gf) =↓ ↑=↓



Warlpiri (2)

The Warlpiri verb specifies a great deal of information about itsarguments.

Lexical entry

panti-rni V (↑ pred)=‘spear〈subj,obj〉’((↑ subj pred)=‘pro’)(↑ subj case)=ERG((↑ obj pred)=‘pro’)(↑ obj case)=ABS

ngarrka-ngku V (↑ pred)=‘man’(↑ case)=ERG

(24)ngarrka-ngku ka wawirri panti-rniman-ERG PRES kangaroo.ABS spear-NONPAST‘The man is spearing the kangaroo.’



Warlpiri: An example

Example

P129. Lexical-Functional Grammar.On whiteboard



Summary

Quote

Semantic roles, syntactic constituents, and gram-matical functions belong to parallel information struc-tures of very different formal character. They are re-lated not by proof-theoretic derivation but by structuralcorrespondences, as a melody is related to the words ofa song. The song is decomposable into parallel melodicand linguistic structures, which jointly constrain thenature of the whole. In the same way, the sentencesof human language are themselves decomposable intoparallel systems of constraints—structural, functional,semantic, and prosodic—which the whole must jointlysatisfy.

Joan Bresnan



Summary

Main ideas

I LFG is a nontransformational theory.I The idea is fairly radical when first proposed.

I A formal system to model human speech: fits in thetradition of generative grammar.

I Psychological plausibility: represent a native speaker’ssyntactic knowledge appropriately.

I Strong typological (universal) basis: analyses shouldcapture cross-linguistic similarities.

LFG brings scholars from different fields together:

I Theoretical linguists

I Descriptive, typological linguists

I Computational linguistics



Summary

LFG is lexical

I LFG assumes that words and lexical items are as importantin providing grammatical information as syntactic elements

I The lexicon is richly structured to capture linguisticgeneralizations.

LFG is functional

Functional syntactic concepts like subject and object are relevantfor the analysis of every language.

I Grammatical functions like subject and object areprimitives of the theory.



Design Principles

Principle I: Variability

External structures (modelled by LFG c-structures) vary acrosslanguages.

Principle II: Universality

Internal structures (modelled by LFG f-structures) are largely in-variant across languages.

Principle III: Monotonicity

The mapping from c-structure to f-structure is not one-to-one,but it is monotonic (information-preserving).



Reading

I Chap. 1, 2, 5. Lexical Functional Grammar

* Chap. 4. Lexical Functional Grammar

* Chapter 13. Grammatical theory: From transformationalgrammar to constraint-based approaches


Documents

Lexical-Functional Grammar - University of Cambridgews390/course/syntax/lfgarch.pdf · 2020. 2. 5. · Lexical-Functional Grammar The idea that words and phrases are alternative means