Lexical Functional Grammar - University of Essex · Nontransformational, constraint-based theories...
Transcript of Lexical Functional Grammar - University of Essex · Nontransformational, constraint-based theories...
Lexical Functional Grammar – 1 / 80
Lexical Functional Grammar
Mary DalrympleCentre for Linguistics and Philology
Oxford University
York Frameworks, 4 May 2010
L F G The constraint-based approach
Lexical Functional Grammar – 2 / 80
Nontransformational, constraint-based theories (LexicalFunctional Grammar, Head-Driven Phrase Structure Grammar,Construction Grammar, Simpler Syntax ...):
■ Different aspects of linguistic structure are realised bydifferent but related linguistic representations.Movement/transformations do not play a role.
L F G
Lexical Functional Grammar – 3 / 80
■ “Semantic roles, syntactic constituents, and grammaticalfunctions belong to parallel information structures of verydifferent formal character. They are related not byproof-theoretic derivation but by structuralcorrespondences, as a melody is related to the words of asong. The song is decomposable into parallel melodic andlinguistic structures, which jointly constrain the nature ofthe whole. In the same way, the sentences of humanlanguage are themselves decomposable into parallel systemsof constraints – structural, functional, semantic, andprosodic – which the whole must jointly satisfy.” (Bresnan,1990)
What theoretical architecture best reflects this view?
L F G Theories and frameworks
Lexical Functional Grammar – 4 / 80
Formal linguistic framework: A set of linguistic objects, rules,and/or processes, and a formal vocabulary for talking aboutthem. Example: X-bar theory: phrase structure rules and trees.
L F G Theories and frameworks
Lexical Functional Grammar – 4 / 80
Formal linguistic framework: A set of linguistic objects, rules,and/or processes, and a formal vocabulary for talking aboutthem. Example: X-bar theory: phrase structure rules and trees.
■ Formally explicit: Provides a way of making systematic,clear, and testable claims about phrase structure.
L F G Theories and frameworks
Lexical Functional Grammar – 4 / 80
Formal linguistic framework: A set of linguistic objects, rules,and/or processes, and a formal vocabulary for talking aboutthem. Example: X-bar theory: phrase structure rules and trees.
■ Formally explicit: Provides a way of making systematic,clear, and testable claims about phrase structure.
■ Embodies some assumptions about how language works:phrases (like VP) have heads (like V),
L F G Theories and frameworks
Lexical Functional Grammar – 4 / 80
Formal linguistic framework: A set of linguistic objects, rules,and/or processes, and a formal vocabulary for talking aboutthem. Example: X-bar theory: phrase structure rules and trees.
■ Formally explicit: Provides a way of making systematic,clear, and testable claims about phrase structure.
■ Embodies some assumptions about how language works:phrases (like VP) have heads (like V),
■ but general enough to encompass a range of differenttheories of phrase structure.
L F G Theories and frameworks
Lexical Functional Grammar – 5 / 80
Linguistic theory: A set of claims about the structure oflanguage(s), which may (or may not) be stated with reference toa particular formal framework.
L F G Theories and frameworks
Lexical Functional Grammar – 5 / 80
Linguistic theory: A set of claims about the structure oflanguage(s), which may (or may not) be stated with reference toa particular formal framework.
■ Example: The claim that all maximal X-bar projectionshave bar level 2 (there is no N′′′ or V′′′′′).
L F G Theories and frameworks
Lexical Functional Grammar – 5 / 80
Linguistic theory: A set of claims about the structure oflanguage(s), which may (or may not) be stated with reference toa particular formal framework.
■ Example: The claim that all maximal X-bar projectionshave bar level 2 (there is no N′′′ or V′′′′′).
■ A well-designed formal framework guides development oftheory by providing explicit representations and theoreticalvocabulary, and aids the linguist in developing betterintuitions about language and (hence) better theories oflinguistic structure.
L F G Theories and frameworks: Other views
Lexical Functional Grammar – 6 / 80
■ Alternative view (NOT LFG): the formal framework shouldnot allow the linguist to formulate rules or describeconstructions that are linguistically impossible.
L F G Theories and frameworks: Other views
Lexical Functional Grammar – 6 / 80
■ Alternative view (NOT LFG): the formal framework shouldnot allow the linguist to formulate rules or describeconstructions that are linguistically impossible.
■ This is a very strong view; e.g. disallows standard phrasestructure rules, since impossible languages can becharacterised with (unconstrained) phrase structure rules(e.g., a language where every sentence is at least 3000words long).
L F G Theories and frameworks: Other views
Lexical Functional Grammar – 6 / 80
■ Alternative view (NOT LFG): the formal framework shouldnot allow the linguist to formulate rules or describeconstructions that are linguistically impossible.
■ This is a very strong view; e.g. disallows standard phrasestructure rules, since impossible languages can becharacterised with (unconstrained) phrase structure rules(e.g., a language where every sentence is at least 3000words long).
■ The LFG view (also HPSG, other constraint-basedtheories): use a simple, clean formal framework, andformulate linguistic theory as a set of claims stated withreference to the framework.
L F G Theories and frameworks: Other views
Lexical Functional Grammar – 6 / 80
■ Alternative view (NOT LFG): the formal framework shouldnot allow the linguist to formulate rules or describeconstructions that are linguistically impossible.
■ This is a very strong view; e.g. disallows standard phrasestructure rules, since impossible languages can becharacterised with (unconstrained) phrase structure rules(e.g., a language where every sentence is at least 3000words long).
■ The LFG view (also HPSG, other constraint-basedtheories): use a simple, clean formal framework, andformulate linguistic theory as a set of claims stated withreference to the framework.
■ Advantage: No need to throw away or reformulate theframework when revisions are needed to the theory.
L F G LFG framework
Lexical Functional Grammar – 7 / 80
Formal framework of LFG:
■ Different aspects of linguistic structure are represented indifferent ways, and are related to one another by piecewisecorrespondence (parts of one structure are related to partsof another structure).
L F G LFG framework
Lexical Functional Grammar – 7 / 80
Formal framework of LFG:
■ Different aspects of linguistic structure are represented indifferent ways, and are related to one another by piecewisecorrespondence (parts of one structure are related to partsof another structure).
■ The core of the formal framework of LFG has remainedremarkably stable since its beginnings in the late 1970s.
L F G LFG framework
Lexical Functional Grammar – 7 / 80
Formal framework of LFG:
■ Different aspects of linguistic structure are represented indifferent ways, and are related to one another by piecewisecorrespondence (parts of one structure are related to partsof another structure).
■ The core of the formal framework of LFG has remainedremarkably stable since its beginnings in the late 1970s.
■ LFG-based theories of linguistic phenomena have evolvedsubstantially since that time, and continue to evolve as newareas are explored and new theoretical proposals areformulated and evaluated.
L F G LFG
Lexical Functional Grammar – 8 / 80
Two aspects of syntactic structure:
■ Functional structure is the abstract functional syntacticorganisation of the sentence, familiar from traditionalgrammatical descriptions, representing syntacticpredicate-argument structure and functional relations likesubject and object.
■ Constituent structure is the overt, more concrete level oflinear and hierarchical organisation of words into phrases.
L F G LFG’s c-structure and f-structure
Lexical Functional Grammar – 9 / 80
IP
NP
N
David
I′
VP
V′
V
greeted
NP
N
Chris
pred ‘greet〈subj,obj〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
L F G C-structure and f-structure
Lexical Functional Grammar – 10 / 80
IP
NP
N
David
I′
VP
V′
V
greeted
NP
N
Chris
pred ‘greet〈subj,obj〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
In GB/Principles and Parameters/Minimalism:
■ C-structure = PF or Spellout?
■ F-structure = S-Structure or LF?
L F G Other linguistic levels
Lexical Functional Grammar – 11 / 80
Since the inception of the theory, there has been much work onother linguistic levels and their relation to c-structure andf-structure:
■ Argument structure and argument linking(Bresnan & Zaenen, 1990; Butt, 1995)
■ The syntax-semantics interface: “glue” semantics(Dalrymple, 1999, 2001; Asudeh, 2004): interestingrelations to categorial approaches, though with differentassumptions about the relation to syntactic structure
■ Information structure and its relation to syntax andsemantics (Butt & King, 2000; Dalrymple & Nikolaeva,2010)
■ Prosodic structure and its relation to syntax and semantics(Mycock, 2006)
L F G LFG as a component of other approaches
Lexical Functional Grammar – 12 / 80
LFG has also been adopted as a component of OT and DOP:
■ OT-LFG: Optimality-theoretic syntax with an LFG base(Bresnan, 2000)
■ LFG-DOP: Data-Oriented Parsing with an LFG base (seehttp://www.nclt.dcu.ie/lfg-dop/publications.html)
L F G F-structure
Lexical Functional Grammar – 13 / 80
What information does functional structure represent?
L F G F-structure
Lexical Functional Grammar – 13 / 80
What information does functional structure represent?
■ Abstract syntactic relations (familiar from traditionalgrammar) like subject, object, adjunct
L F G F-structure
Lexical Functional Grammar – 13 / 80
What information does functional structure represent?
■ Abstract syntactic relations (familiar from traditionalgrammar) like subject, object, adjunct
■ Locus of subcategorisation
L F G F-structure
Lexical Functional Grammar – 13 / 80
What information does functional structure represent?
■ Abstract syntactic relations (familiar from traditionalgrammar) like subject, object, adjunct
■ Locus of subcategorisation
■ Criteria: anaphoric binding patterns, long-distancedependencies, control, honorification, agreement,casemarking, ...
L F G F-structure
Lexical Functional Grammar – 13 / 80
What information does functional structure represent?
■ Abstract syntactic relations (familiar from traditionalgrammar) like subject, object, adjunct
■ Locus of subcategorisation
■ Criteria: anaphoric binding patterns, long-distancedependencies, control, honorification, agreement,casemarking, ...
■ F-structure vocabulary is universal across languages
L F G Functional structure
Lexical Functional Grammar – 14 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
L F G Functional structure
Lexical Functional Grammar – 14 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
■ pred, tense num: attributes
L F G Functional structure
Lexical Functional Grammar – 14 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
■ pred, tense num: attributes
■ ‘go〈subj〉’, David, sg: values
L F G Functional structure
Lexical Functional Grammar – 14 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
■ pred, tense num: attributes
■ ‘go〈subj〉’, David, sg: values
■ past, sg: symbols (a kind of value)
L F G Functional structure
Lexical Functional Grammar – 14 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
■ pred, tense num: attributes
■ ‘go〈subj〉’, David, sg: values
■ past, sg: symbols (a kind of value)
■ ‘boy’, ‘go〈subj〉’: semantic forms
L F G F-structures
Lexical Functional Grammar – 15 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
adj{[
pred ‘quickly’]}
An f-structure can be the value of an attribute. Attributes withf-structure values are the grammatical functions: subj, obj,objθ, comp, xcomp, ...
L F G F-structures
Lexical Functional Grammar – 16 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
num sg
]
adj{[
pred ‘quickly’]}
A set of f-structures can also be a value of an attribute.
L F G Sets of f-structures
Lexical Functional Grammar – 17 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
]
[pred ‘George’
]
adj{[
pred ‘quickly’]}
Sets of f-structures represent:
■ adjuncts (there can be more than one adjunct) or
L F G Sets of f-structures
Lexical Functional Grammar – 17 / 80
pred ‘go〈subj〉’
tense past
subj
[pred ‘David’
]
[pred ‘George’
]
adj{[
pred ‘quickly’]}
Sets of f-structures represent:
■ adjuncts (there can be more than one adjunct) or
■ coordinate structures (there can be more than oneconjunct)
L F G Describing F-structures
Lexical Functional Grammar – 18 / 80
(f num) = sg
is a functional equation.
(f a) = v holds if and only if f is an f-structure, a is a symbol,and the pair 〈a, v〉 ∈ f .
A set of formulas describing an f-structure is a functionaldescription.
L F G More Complex Descriptions
Lexical Functional Grammar – 19 / 80
(f subj num) = (g num) = sg
f
pred ‘go〈subj〉’
subj g
[pred ‘David’
num sg
]
L F G Finding the Right F-structure
Lexical Functional Grammar – 20 / 80
Hindi verbs show person, number, and gender agreement:
Ram
Ramcalegaa
go.future
‘Ram will go.’
Ram (g pred) = ‘Ram’(g case) = nom(g pers) = 3(g num) = sg(g gend) = masc
calegaa (f pred) = ‘go〈subj〉’(f subj case) = nom(f subj pers) = 3(f subj num) = sg(f subj gend) = masc
(f subj) = g
L F G F-description and its solution
Lexical Functional Grammar – 21 / 80
(g pred) = ‘Ram’(g case) = nom(g pers) = 3(g num) = sg(g gend) = masc
(f pred) = ‘go〈subj〉’(f subj) = g
f
pred ‘go〈subj〉’
subj g
pred ‘Ram’
case nom
pers 3
num sg
gend masc
(f subj case) = (g case) = nom(f subj num) = (g num) = sg(f subj pers) = (g pers) = 3(f subj gend) = (g gend) = masc
L F G Formal descriptions: LFG vs HPSG
Lexical Functional Grammar – 22 / 80
■ HPSG takes a different view of formal descriptions fromLFG. The HPSG view goes back to Functional UnificationGrammar (Kay, 1984), where unification (an operation onstructures) was used to combine structures:
■ in HPSG, the constraints look (as much as possible) likethe structures.
■ That is why you sometimes see a set of instructions in whatlooks like a representation – it is actually a constraint ordescription in the (apparent) form of a structure.
L F G Formal descriptions: LFG vs HPSG
Lexical Functional Grammar – 23 / 80
HPSG’s Argument Realisation Principle (Sag et al., 2003, 432):
word:
SYN
VAL
SPR A
COMPS B C
GAP C
ARG-STR A ⊕ B
: list subtraction⊕: list addition
L F G Generalisations and constructions
Lexical Functional Grammar – 24 / 80
■ Expressing generalisations over functional descriptions:templates (Dalrymple et al., 2004; Asudeh et al., 2008)
L F G Generalisations and constructions
Lexical Functional Grammar – 24 / 80
■ Expressing generalisations over functional descriptions:templates (Dalrymple et al., 2004; Asudeh et al., 2008)
■ Templates are names for bundles of functional equationsthat characterise a construction.
L F G Generalisations and constructions
Lexical Functional Grammar – 24 / 80
■ Expressing generalisations over functional descriptions:templates (Dalrymple et al., 2004; Asudeh et al., 2008)
■ Templates are names for bundles of functional equationsthat characterise a construction.
■ Templates can be defined in terms of other templates,giving something like the inheritence hierarchy of HPSG(but involving relations among descriptions rather thanlinguistic objects).
L F G Generalisations and constructions
Lexical Functional Grammar – 24 / 80
■ Expressing generalisations over functional descriptions:templates (Dalrymple et al., 2004; Asudeh et al., 2008)
■ Templates are names for bundles of functional equationsthat characterise a construction.
■ Templates can be defined in terms of other templates,giving something like the inheritence hierarchy of HPSG(but involving relations among descriptions rather thanlinguistic objects).
■ Templates can be associated with words or with units thatare bigger than words, and are used to describeconstructions in the Construction Grammar sense.
L F G Generalisations and constructions
Lexical Functional Grammar – 24 / 80
■ Expressing generalisations over functional descriptions:templates (Dalrymple et al., 2004; Asudeh et al., 2008)
■ Templates are names for bundles of functional equationsthat characterise a construction.
■ Templates can be defined in terms of other templates,giving something like the inheritence hierarchy of HPSG(but involving relations among descriptions rather thanlinguistic objects).
■ Templates can be associated with words or with units thatare bigger than words, and are used to describeconstructions in the Construction Grammar sense.
■ This is a relatively recent area of exploration in LFG.
L F G Semantic Forms
Lexical Functional Grammar – 25 / 80
Subcategorisation requirements are imposed at f-structure (notc-structure) – a predicate specifies a set of grammaticalfunctions, and the phrase structure grammar of the languagedetermines where in the tree these functions can appear.Subcategorisation requirements are specified by semantic forms:
(f pred) = ‘go〈subj〉’
Semantic forms have argument lists that list the arguments theyrequire.
L F G Grammatical functions
Lexical Functional Grammar – 26 / 80
Non-argument topic Discourse functionfocus
Argument Core subj(governable) obj Non-discourse function
objθNon-core oblθ
compNon-argument adj(unct)
(from Borjars & Vincent 2004)
L F G Completeness
Lexical Functional Grammar – 27 / 80
Completeness requires: All arguments which are listed in thesemantic form must be present.
(f pred) = ‘go〈subj〉’
“Go” must have a subj.
L F G Coherence
Lexical Functional Grammar – 28 / 80
Coherence requires: No arguments which are not listed in thesemantic form may be present.
(f pred) = ‘go〈subj〉’
“Go” may not have a obj.
L F G Coherence
Lexical Functional Grammar – 28 / 80
Coherence requires: No arguments which are not listed in thesemantic form may be present.
(f pred) = ‘go〈subj〉’
“Go” may not have a obj.
Completeness and coherence are the equivalent (more or less) ofthe Theta Criterion of GB theory, or the Valence Principle andRoot Condition of HPSG.
L F G Semantic Forms and Uniqueness
Lexical Functional Grammar – 29 / 80
*watiman.abs
ka
presparnka-mi
run-nonpastkarnta
woman.abs
‘The man runs the woman.’ (Warlpiri)
wati (g pred) = ‘man’karnta (g pred) = ‘woman’
Each use of a semantic form is unique.
L F G Conflicting Semantic Forms
Lexical Functional Grammar – 30 / 80
wati (g pred) = ‘man’karnta (g pred) = ‘woman’
Ill-formed f-structure:
pred ‘run〈subj〉’
tense pres
subj g[pred ‘man’/‘woman’
]
L F G Optionality
Lexical Functional Grammar – 31 / 80
njuchi
beeszi-na-lum-a
subj-past-bite-indicativealenje
hunters
‘The bees bit the hunters.’ (Chichewa)
zi-na-lum-a
subj-past-bite-indicativealenje
hunters
‘They bit the hunters.’
zi-na-lum-a: ((f subj pred) = ‘pro’)
zi-na-lum-a optionally contributes a pred for its subj.
L F G Overt subject
Lexical Functional Grammar – 32 / 80
njuchi
beeszi-na-lum-a
subj-past-bite-indicativealenje
hunters
‘The bees bit the hunters.’
f
pred ‘bite〈subj,obj〉’
subj
[pred ‘bees’
nounclass 10
]
obj
[pred ‘hunters’
nounclass 2
]
L F G No overt subject
Lexical Functional Grammar – 33 / 80
zi-na-lum-a
subj-past-bite-indicativealenje
hunters
‘They bit the hunters.’
f
pred ‘bite〈subj,obj〉’
subj
[pred ‘pro’
nounclass 10
]
obj
[pred ‘hunters’
nounclass 2
]
L F G Optionality: Clitics
Lexical Functional Grammar – 34 / 80
Juan
Juanvio
sawa
prepPedro.
Pedro
‘Juan saw Pedro.’ (Spanish)
Juan
Juanlo
acc.masc.sg.cliticvio.
saw
‘Juan saw him.’
Juan
Juanlo
acc.masc.sg.cliticvio
sawa
prepPedro.
Pedro
‘Juan saw Pedro.’
L F G Optionality: Clitics
Lexical Functional Grammar – 35 / 80
Pedro (f pred) = ‘Pedro’(f gend) = masc(f num) = sg
lo ((f pred) = ‘pro’)(f gend) = masc(f num) = sg
L F G Optionality: Clitics
Lexical Functional Grammar – 35 / 80
Pedro (f pred) = ‘Pedro’(f gend) = masc(f num) = sg
lo ((f pred) = ‘pro’)(f gend) = masc(f num) = sg
lo optionally contributes a pred.
L F G Optionality: Clitics
Lexical Functional Grammar – 36 / 80
Juan
Juanlo
acc.masc.sg.cliticvio
sawa
prepPedro.
Pedro
‘Juan saw Pedro.’
L F G Optionality: Clitics
Lexical Functional Grammar – 36 / 80
Juan
Juanlo
acc.masc.sg.cliticvio
sawa
prepPedro.
Pedro
‘Juan saw Pedro.’
pred ‘see〈subj,obj〉’
subj
pred ‘Juan’
gend masc
num sg
obj f
pred ‘Pedro’
gend masc
num sg
L F G Optionality and clitic doubling
Lexical Functional Grammar – 37 / 80
Juan
Juanlo
acc.masc.sg.cliticvio.
saw
‘Juan saw him.’
pred ‘see〈subj,obj〉’
subj
pred ‘Juan’
gend masc
num sg
obj f
pred ‘pro’
gend masc
num sg
L F G C-structure and f-structure
Lexical Functional Grammar – 38 / 80
IP
NP
N
David
I′
VP
V′
V
greeted
NP
N
Chris
pred ‘greet〈subj,obj〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
L F G Motivating Constituent Structure
Lexical Functional Grammar – 39 / 80
What information does constituent structure represent?
L F G Motivating Constituent Structure
Lexical Functional Grammar – 39 / 80
What information does constituent structure represent?
■ Represents hierarchical phrasal groupings
L F G Motivating Constituent Structure
Lexical Functional Grammar – 39 / 80
What information does constituent structure represent?
■ Represents hierarchical phrasal groupings
■ Criteria depend on surface syntactic properties, notsemantic intuitions or facts about abstract functionalsyntactic structure
L F G Motivating Constituent Structure
Lexical Functional Grammar – 39 / 80
What information does constituent structure represent?
■ Represents hierarchical phrasal groupings
■ Criteria depend on surface syntactic properties, notsemantic intuitions or facts about abstract functionalsyntactic structure
■ Varies greatly across languages
L F G Constituent Structure
Lexical Functional Grammar – 40 / 80
■ Some theories (GB/Principles and Parameters, NOT LFG):Subjects always appear in the specifier of IP.
L F G Constituent Structure
Lexical Functional Grammar – 40 / 80
■ Some theories (GB/Principles and Parameters, NOT LFG):Subjects always appear in the specifier of IP.
■ LFG does not assume that subjects are defined in terms ofphrase structure position, or that subjects must alwaysappear in a particular position in the tree.
L F G Constituent Structure
Lexical Functional Grammar – 40 / 80
■ Some theories (GB/Principles and Parameters, NOT LFG):Subjects always appear in the specifier of IP.
■ LFG does not assume that subjects are defined in terms ofphrase structure position, or that subjects must alwaysappear in a particular position in the tree.
■ However, there are structure-function mappinggeneralisations which state that phrases with particularfunctions tend to appear in particular phrase structurepositions.
L F G Constituent Structure
Lexical Functional Grammar – 40 / 80
■ Some theories (GB/Principles and Parameters, NOT LFG):Subjects always appear in the specifier of IP.
■ LFG does not assume that subjects are defined in terms ofphrase structure position, or that subjects must alwaysappear in a particular position in the tree.
■ However, there are structure-function mappinggeneralisations which state that phrases with particularfunctions tend to appear in particular phrase structurepositions.
■ In English, the specifier of IP is associated with the subjectfunction; in other languages, it is associated with TOPIC orFOCUS. More below.
L F G Lexical Integrity
Lexical Functional Grammar – 41 / 80
Lexical Integrity (Bresnan, 1982): Morphologically completewords are leaves of the c-structure tree, and each leafcorresponds to one and only one c-structure node.
L F G Lexical Integrity
Lexical Functional Grammar – 41 / 80
Lexical Integrity (Bresnan, 1982): Morphologically completewords are leaves of the c-structure tree, and each leafcorresponds to one and only one c-structure node.
English: cause to run
Japanese: hasirasetarun.caus.past
pred ‘cause〈subj,obj,xcomp〉’
subj [ ]
obj [ ]
xcomp
pred ‘run〈subj〉’
subj
Words in one language can express the same f-structure asphrases in another language: Lexical Integrity holds atc-structure, not f-structure.
L F G Economy of Expression
Lexical Functional Grammar – 42 / 80
Economy of Expression (Bresnan, 2001): All syntactic phrasestructure nodes are optional, and are not used unless required byindependent principles (completeness, coherence, semanticexpressivity).
CP
C′
C
Is
IP
NP
N
David
I′
VP
V
yawning
L F G
Lexical Functional Grammar – 43 / 80
CP
NP
N
kogda
when
C′
IP
I′
I
rodilsja
born
VP
NP
N
Lermontov
Lermontov
‘When was Lermontov born?’
L F G C-structure and f-structure
Lexical Functional Grammar – 44 / 80
IP
NP
N
David
I′
VP
V′
V
greeted
NP
N
Chris
pred ‘greet〈subj,obj〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
L F G C- and F-Structure
Lexical Functional Grammar – 45 / 80
V
greeted
[pred ‘greet〈subj,obj〉’
tense past
]φ
φ function relates c-structure nodes to f-structures.
(Function: Every c-structure node corresponds to exactly onef-structure.)
L F G Many Corresponding Nodes
Lexical Functional Grammar – 46 / 80
VP
V′
V
greeted
[pred ‘greet〈subj, obj〉’
tense past
]φ
Many c-structure nodes can correspond to the same f-structure.
L F G No Corresponding Node
Lexical Functional Grammar – 47 / 80
S
V
kowareta
break.past
pred ‘break〈subj〉’
tense past
subj[pred ‘pro’
]
φ
Some f-structures have no corresponding c-structure node.
L F G No Corresponding Node
Lexical Functional Grammar – 47 / 80
S
V
kowareta
break.past
pred ‘break〈subj〉’
tense past
subj[pred ‘pro’
]
φ
Some f-structures have no corresponding c-structure node.
These are formal, mathematical facts about thec-structure/f-structure relation. What are the linguistic facts?
L F G Mapping regularities
Lexical Functional Grammar – 48 / 80
C-structure heads are f-structure heads:
VP
V′
V
greeted
[pred ‘greet〈subj, obj〉’
tense past
]φ
L F G Mapping Regularities
Lexical Functional Grammar – 49 / 80
Specifiers are filled by grammaticized discourse functions SUBJ,TOPIC, FOCUS.
L F G Mapping Regularities
Lexical Functional Grammar – 49 / 80
Specifiers are filled by grammaticized discourse functions SUBJ,TOPIC, FOCUS.
Specifier of IP in English: SUBJ
IP
NP
N
David
I′
VP
V
yawned
pred ‘yawn〈subj〉’
subj[pred ‘David’
]
L F G Mapping regularities
Lexical Functional Grammar – 50 / 80
Specifier of IP in Russian: Topic or Focus
IP
NP
Evgenija Onegina
Eugene Onegin
I′
I
napisal
wrote
VP
NP
N
Puskin
Pushkin
pred ‘write〈subj,obj〉’
topic
{[pred
‘EugeneOnegin’
]}
subj[pred ‘Pushkin’
]
obj
L F G Mapping regularities
Lexical Functional Grammar – 51 / 80
Specifier of IP in Bulgarian: Focus; Specifier of CP: Topic
CP
NP
N
Ivan
Ivan
C′
IP
NP
N
kakvo
what
I′
I
pravi
does
pred ‘do〈subj,obj〉’
topic[pred ‘Ivan’
]
subj
focus[pred ‘what’
]
obj
L F G Mapping regularities
Lexical Functional Grammar – 52 / 80
Specifier of CP in English: Focus
CP
NP
N
What
C′
C
is
IP
NP
N
David
I′
VP
V
eating
pred ‘eat〈subj,obj〉’
focus[pred ‘what’
]
subj[pred ‘David’
]
obj
L F G Mapping regularities
Lexical Functional Grammar – 53 / 80
Specifier of CP in Finnish: Focus
CP
NP
N
Mikolta
Mikko.abl
C′
IP
NP
N
Anna
Anna
I′
I
sai
got
VP
NP
N
kukkia
flowers.part
pred ‘get〈subj,obj,oblsource〉’
focus[pred ‘Mikko’
]
oblsource
topic[pred ‘Anna’
]
subj
obj[pred ‘flowers’
]
L F G Complements: Functional Categories
Lexical Functional Grammar – 54 / 80
Complement of functional category is f-structure co-head:
IP
NP
N
David
I′
I
is
VP
V
yawning
pred ‘yawn〈subj〉’
subj[pred ‘David’
]
L F G Complements: Functional Categories
Lexical Functional Grammar – 55 / 80
IP
NP
N
Anna
Anna
I′
I
budet
future
VP
V′
V
citat’
read.inf
NP
N
knigu
book
pred ‘read〈subj,obj〉’
tense future
topic{[
pred ‘Anna’]}
subj
obj[pred ‘book’
]
L F G Complements of Lexical Categories
Lexical Functional Grammar – 56 / 80
Complement of lexical category is f-structure complement(non-subject argument):
IP
NP
N
David
I′
VP
V′
V
greeted
NP
N
Chris
pred ‘greet〈subj,obj〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
L F G Complements of Lexical Categories
Lexical Functional Grammar – 57 / 80
IP
NP
N
David
I′
VP
V′
V
gave
NP
N
Chris
NP
Det
a
N′
N
book
pred ‘give〈subj,obj,objtheme〉’
subj[pred ‘David’
]
obj[pred ‘Chris’
]
objtheme
spec
[pred ‘a’
]
pred ‘book’
L F GConstraining the c-structure/f-structurecorrespondence
Lexical Functional Grammar – 58 / 80
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
L F GConstraining the c-structure/f-structurecorrespondence
Lexical Functional Grammar – 58 / 80
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
V′ −→ V
L F G Local F-Structure Reference
Lexical Functional Grammar – 59 / 80
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
V′ −→ V
the current c-structure node (“self”): ∗the immediately dominating node (“mother”): ∗
the c-structure to f-structure function: φ
L F G Rule Annotation
Lexical Functional Grammar – 60 / 80
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
L F G Rule Annotation
Lexical Functional Grammar – 60 / 80
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
V′ −→ Vφ(∗) = φ(∗)
mother’s (V′’s) f-structure = self’s (V’s) f-structure
L F G Simplifying the Notation
Lexical Functional Grammar – 61 / 80
φ(∗) (mother’s f-structure) = ↑φ(∗) (self’s f-structure) = ↓
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
L F G Simplifying the Notation
Lexical Functional Grammar – 61 / 80
φ(∗) (mother’s f-structure) = ↑φ(∗) (self’s f-structure) = ↓
V′
V
yawned
[pred ‘yawn〈subj〉’
tense past
]φ
V′ −→ V↑= ↓
mother’s f-structure = self’s f-structure
L F G Using the Notation
Lexical Functional Grammar – 62 / 80
V′ −→ V↑= ↓
mother’s f-structure = self’s f-structure
L F G Using the Notation
Lexical Functional Grammar – 62 / 80
V′ −→ V↑= ↓
mother’s f-structure = self’s f-structure
V′
V↑ = ↓
L F G Using the Notation
Lexical Functional Grammar – 62 / 80
V′ −→ V↑= ↓
mother’s f-structure = self’s f-structure
V′
V↑ = ↓
L F G Using the Notation
Lexical Functional Grammar – 62 / 80
V′ −→ V↑= ↓
mother’s f-structure = self’s f-structure
V′
V↑ = ↓
[ ]
L F G More rules
Lexical Functional Grammar – 63 / 80
V′ −→ Vφ(∗) = φ(∗)
NP(φ(∗) obj) = φ(∗)
mother’s f-structure’s obj = self’s f-structure
In simpler form:
V′ −→ V↑= ↓
NP(↑ obj) = ↓
L F G Using the Notation
Lexical Functional Grammar – 64 / 80
V′ −→ V↑= ↓
NP(↑ obj) = ↓
L F G Using the Notation
Lexical Functional Grammar – 64 / 80
V′ −→ V↑= ↓
NP(↑ obj) = ↓
V′
V NP
[obj [ ]
]
L F G Terminal nodes
Lexical Functional Grammar – 65 / 80
V
yawned
[pred ‘yawn〈subj〉’
tense past
]
L F G Terminal nodes
Lexical Functional Grammar – 65 / 80
V
yawned
[pred ‘yawn〈subj〉’
tense past
]
Expressible as:
V −→ yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Terminal nodes
Lexical Functional Grammar – 65 / 80
V
yawned
[pred ‘yawn〈subj〉’
tense past
]
Expressible as:
V −→ yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
Standard form:
yawned V (↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Phrase structure rules: English
Lexical Functional Grammar – 66 / 80
IP −→(
NP(↑ subj) = ↓
) (I′
↑= ↓
)
I′ −→(
I↑= ↓
) (VP↑= ↓
)
VP −→(
V↑= ↓
)
NP −→(
N↑= ↓
)
L F G Lexical entries: English
Lexical Functional Grammar – 67 / 80
yawned V (↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
David N (↑ pred) = ‘David’
L F G Lexical entries: English
Lexical Functional Grammar – 67 / 80
yawned V (↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
David N (↑ pred) = ‘David’
(Standard LFG practice: include only features relevant foranalysis under discussion.)
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(↑ subj) = ↓
N↑ = ↓
David
(↑ pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
V↑ = ↓
yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(↑ subj) = ↓
N↑ = ↓
David
(fn pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
V↑ = ↓
yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(↑ subj) = ↓
Nfnp = fn
David
(fn pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
V↑ = ↓
yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
V↑ = ↓
yawned
(↑ pred) = ‘yawn〈subj〉’(↑ tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
V↑ = ↓
yawned
(fv pred) = ‘yawn〈subj〉’(fv tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
↑ = ↓
VP↑ = ↓
Vfvp = fv
yawned
(fv pred) = ‘yawn〈subj〉’(fv tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
↑ = ↓
VPfi′ = fvp
Vfvp = fv
yawned
(fv pred) = ‘yawn〈subj〉’(fv tense) = past
L F G Analysis: English
Lexical Functional Grammar – 68 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
fip = fi′
VPfi′ = fvp
Vfvp = fv
yawned
(fv pred) = ‘yawn〈subj〉’(fv tense) = past
L F G Solving the Description
Lexical Functional Grammar – 69 / 80
(fip subj) = fnpfnp = fn(fn pred) = ‘David’fip = fi′
fi′ = fvpfvp = fv(fv pred) = ‘yawn〈subj〉’(fv tense) = past
fipfi′
fvpfv
pred ‘yawn〈subj〉’
tense past
subjfnpfn
[pred ‘David’
]
L F G Final result
Lexical Functional Grammar – 70 / 80
IP
NP(fip subj) = fnp
Nfnp = fn
David
(fn pred) = ‘David’
I′
fip = fi′
VPfi′ = fvp
Vfvp = fv
yawned
(fv pred) = ‘yawn〈subj〉’(fv tense) = past
pred ‘yawn〈subj〉’
tense past
subj[pred ‘David’
]
L F G Warlpiri
Lexical Functional Grammar – 71 / 80
gf ≡ {subj | obj | oblθ}
IP −→
NP(↑ focus) = ↓(↑ gf) = ↓
(I′
↑= ↓
)
I′ −→(
I↑= ↓
) (S
↑= ↓
)
S −→ { NP(↑ gf) = ↓
| V↑= ↓
}∗
L F G Warlpiri verbs
Lexical Functional Grammar – 72 / 80
panti-rni V (↑ pred) = ‘spear〈subj,obj〉’((↑ subj pred) = ‘pro’)(↑ subj case) = erg((↑ obj pred) = ‘pro’)(↑ obj case) = abs
L F G Warlpiri
Lexical Functional Grammar – 73 / 80
IP
NP(↑ focus)=↓(↑ gf)=↓
N↑= ↓
ngarrka-ngku
man-erg
(↑ pred) = ‘man’(↑ case) = erg
I′
↑=↓
I↑=↓
ka
pres
S↑=↓
NP(↑ gf)=↓
N↑= ↓
wawirri
kangaroo.abs
(↑ pred) = ‘kangaroo’(↑ case) = abs
V↑=↓
panti-rni
spear-nonpast
(↑ pred) = ‘spear〈subj,obj〉’((↑ subj pred) = ‘pro’)(↑ subj case) = erg
((↑ obj pred) = ‘pro’)(↑ obj case) = abs
pred ‘spear〈subj,obj〉’
focus
[pred ‘man’
case erg
]
subj
obj
[pred ‘kangaroo’
case abs
]
L F G Chichewa
Lexical Functional Grammar – 74 / 80
S −→(
NP(↑ subj) = ↓
),
(NP
(↑ topic) = ↓
),
(VP
↑= ↓
)
VP −→(
V′
↑ = ↓
)
V′ −→(
V↑ = ↓
) (NP
(↑ obj) = ↓
)
Comma between daughters in S rule: daughters of S areunordered
L F G Chichewa verbs
Lexical Functional Grammar – 75 / 80
zi-na-wa-lum-a V (↑ pred) = ‘bite〈subj,obj〉’((↑ subj pred) = ‘pro’)(↑ subj nounclass) = 10(↑ obj pred) = ‘pro’(↑ obj nounclass) = 2
L F G Chichewa
Lexical Functional Grammar – 76 / 80
S
NP(↑ subj)=↓
njuchi
bees
(↑ pred) = ‘bees’(↑ nounclass) = 10
VP↑=↓
V′
↑=↓
V
zi-na-wa-lum-a
subj-past-obj-bite-indicative
(↑ pred) = ‘bite〈subj,obj〉’((↑ subj pred) = ‘pro’)(↑ subj nounclass) = 10(↑ obj pred) = ‘pro’(↑ obj nounclass) = 2
pred ‘bite〈subj,obj〉’
subj
[pred ‘bees’
nounclass 10
]
obj
[pred ‘pro’
nounclass 2
]
L F G Chichewa
Lexical Functional Grammar – 77 / 80
S
VP↑=↓
V′
↑=↓
V
zi-na-lum-a
subj-past-obj-bite-indicative
(↑ pred) = ‘bite〈subj,obj〉’((↑ subj pred) = ‘pro’)(↑ subj nounclass) = 10
NP(↑ obj)=↓
alenje
hunters
(↑ pred) = ‘hunter’(↑ nounclass) = 10
pred ‘bite〈subj,obj〉’
subj
[pred ‘pro’
nounclass 10
]
obj
[pred ‘hunter’
nounclass 2
]
L F G For more information
Lexical Functional Grammar – 78 / 80
■ For more on LFG, visit the LFG website:http://www.essex.ac.uk/linguistics/LFG/
■ Introductions to LFG: Bresnan (2001), Dalrymple (2001),Falk (2001)
■ SOAS, Essex, and Oxford hold student-oriented meetingseach term for discussion of issues in LFG, including studentpresentations:http://se-lfg.tk/
L F G Bibliography
Lexical Functional Grammar – 79 / 80
Lexical Functional Gramma
References
Asudeh, Ash. 2004. Resumption as Resource Management.Ph.D. thesis, Stanford University.
Asudeh, Ash, Mary Dalrymple, & Ida Toivonen. 2008.Constructions with lexical integrity: Templates as thelexicon-syntax interface. In Miriam Butt & Tracy HollowayKing(editors), On-line Proceedings of the LFG2007 Conference. URLhttp://csli-publications.stanford.edu/LFG/13/lfg08.htm
Borjars, Kersti & Nigel Vincent. 2004. Introduction to LFG.Slides from the Winter School in LFG and ComputationalLinguistics, University of Canterbury.
Bresnan, Joan. 1982. The passive in lexical theory. In JoanBresnan (editor), The Mental Representation of Grammatical
Relations, pp. 3–86. Cambridge, MA: The MIT Press.