Head-Driven Phrase Structure Grammar (HPSG)cl.indiana.edu/~md7/15/614/slides/06-hpsg/06-hpsg-1x3.pdfIntroduction Signs Phrases Raising/Equi UDCs Head-Driven Phrase Structure Grammar

Introduction Signs Phrases Raising/Equi UDCs

Head-Driven Phrase Structure Grammar (HPSG)

Linguistics 614

With thanks to Detmar Meurers

Spring 2015


Motivation

Head-Driven Phrase Structure Grammar (HPSG)

The main mechanism of HPSG for linguistic analysis lies in itsfeature architecture

Trees, bar-level distinctions, subcategorization, agreement,etc. are handled via feature structures

Rough comparison to LFG:

Similar: use of feature structures, highly lexical

Different: only one layer of syntax, functions are notprimitives, stronger reliance on structure-sharing


Motivation

Preview: lexical entry for letter

wordphon <letter>

syn

head

noun

agr 1

[pernum 3sng

gend neut

]

spr

⟨D

[agr 1

count +

]⟩

comps⟨(PP

)⟩

(This entry more or less follows Carnie (2013) / Sag & Wasow (1999),

which most of the rest of the slides don’t.)


Motivation

Unpacking letter

syn: holds syntactic information (vs. phon or sem)

synsem is also commonly used (e.g., these slides)

head: category & inflectional properties

spr & comps: subcategorization properties

Sometimes 3 lists (subj, spr, comps), sometimes 1 list(subcat)Note that detailed properties of selected items can be specified


Components: from a linguistic perspective

HPSG grammarsLinguistic perspective

An HPSG grammar, from a linguistic perspective, consists of

a) a lexicon: licensing basic words

b) lexical rules: licensing derived words

c) immediate dominance (ID) schemata: licensingconstituent structure

d) linear precedence (LP) statements: constraining wordorder

e) a set of grammatical principles: expressing generalizationsabout linguistic objects



Lexical rules

e.g., plural rule (adapted from Carnie 2013):

noun

phon 1

arg-st⟨[count +

]⟩

⇒

word

phon FNPL

(1)

syn|head|agr|num pl

Any feature not specified is unchanged



Grammatical principles

Grammatical principles allow one to generalize over entire classesof objects

e.g., every complete sentence in English is tensed

These can be interpreted as constraints on the acceptable featurestructures

HPSG is constraint-based: grammars are sets of constraintswhich hold simultaneously

This determines the collections of admissible linguisticstructures

Constraints do not define an order of the derivation orgeneration of linguistic signs


Components: from a formal perspective

HPSG grammarsFormal perspective

An HPSG grammar formally consists of

I. the signature as declaration of the domain, and

II. the theory constraining the domain.



The signature of an HPSG grammar

The signature

defines the ontology (‘declaration of what exists’):

which kind of objects are distinguished, andwhich properties of which objects are modeled.

consists of

the type hierarchy (or sort hierarchy) andthe appropriateness conditions, defining which type haswhich appropriate attributes (or features) with whichappropriate values.



Signature: some examples

Signs

word phrase

signphon list(phonstring)

synsem synsem

Part of speech

adj adv det noun prep verb

part-of-speech



Models of linguistic objects

What do the mathematical structures used as model for HPSGtheories look like?

The objects are modeled by typed feature structures, whichcan be notated as directed graphs.

These models represent objects in the world → they are totalwith respect to the ontology declared in the signature.

Naturally, knowledge about objects in the world is partial.

Formally speaking, the feature structures used as models are

totally well-typed: Every type has every one of the attributesand their values which are appropriate for it.sort-resolved: Every type is maximally specific.

type and sort as well as attribute and feature are used synonymously.



An example for a typed feature structure



The theory of an HPSG grammar

A theory is a set of description language statements, oftenreferred to as the constraints.

The theory singles out a subset of the objects declared in thesignature, namely those which are grammatical.

A linguistic object is admissible with respect to a theory iff itsatisfies each of the descriptions in the theory and so doeseach of its substructures.



How do we express a theory? Descriptions

A description language describes sets of objects.

The AVM (attribute value matrix) notation is used tocompactly write down these descriptions.

Descriptions consists of three building blocks:

Type descriptions single out all objects of a given type

Example: wordAttribute-value pairs describe objects that have a particularproperty.

The attribute must be appropriate for the particular type ofobject, and the value can be any kind of description.

Example:

[spouse

[name mary

]]

Tags (structure sharing) to specify token identity

Example: 1



Descriptions (cont.)

Complex descriptions are obtained through

conjunction (∧)disjunction (∨)negation (¬)

In the AVM notation, conjunction is implicit.



An example AVM - The pronoun she

wordphon <she>

synsem|loc

local

cat

cat

head

[noun

case nom

]

subcat 〈〉

cont

ppro

index 1

ref

per third

num sing

gend fem

restr {}

context

context

backgr

psoa

reln female

inst 1


Aspects of the signature

SignsAspects of the signature

word[phrasedtrs constituent-structure

]

signphon list(phonstring)synsem synsem

synsemlocal localnon-local non-local

localcategory categorycontent contentcontext context

categoryhead headsubcat list(synsem)marking marking



Motivating subcat

(1) a. I laugh. (<NP>)

b. I saw him. (<NP NP>)

c. I give her the book. (<NP NP NP>)

d. I said that she left. (<NP S[that]>)

Cannot always be derived from semantics:

(2) a. Paul ate a steak. (<NP>)

b. Paul ate. (<NP NP>)

(3) a. Paul devoured a steak. (<NP>)

b. * Paul devoured (<NP NP>)



Syntactic category information

marker determiner

[functionalspec synsem

]

adjective

verbvform vformaux booleaninv boolean

[nouncase case

][prepositionpform pform

]. . .

substantiveprd booleanmod mod-synsem

head



Properties of particular categories

finite infinitive base gerund present-part. past-part. passive-part.

vform

nominative accusative

case

of to . . .

pform



Motivating VFORM

(4) a. Peter will win the race. (base form)

b. * Peter will won the race.

c. * Peter will to win the race.

(5) a. Peter has won the race. (past participle)

b. * Peter has win the race.

c. Peter has to win the race.(→ different verb)

(6) a. Peter seems to win the race. (to-infinitive)

b. * Peter seems win the race.

c. * Peter seems won the race.



Motivating CASE

(7) a. He left. (nom)

b. * Him left.

(8) a. She sees him. (acc)

b. * She sees he.



Indices

referential there it

indexperson personnumber numbergender gender

first second third

person

singular plural

number

masculine feminine neuter

gender



Semantic representations

quant

[laugh’laugher ref

]

give’giver refgiven refgift ref

think’thinker refthought psoa

psoa nom-objindex indexrestriction set(psoa)

content



Auxiliary data structures

true false

boolean

empty-listnon-empty-listfirst ⊤rest list

list . . .

⊤

Alternative names for the attributes first (ft) and rest (rt) ofnon-empty-list are head (hd) and tail (tl).



Abbreviations for describing lists

empty-list is abbreviated as e-list, <>

non-empty-list is abbreviated as ne-list

[first 1rest 2

]is abbreviated as

⟨1 | 2

⟩

⟨. . . 1 | 〈〉

⟩is abbreviated as

⟨. . . 1

⟩

first 1

rest

[first 2rest 3

] is abbreviated as

⟨1 , 2 | 3

⟩



Abbreviations for common AVMs

Pollard and Sag (1994) make use of the following abbreviations fordescribing synsem objects:

Abbrev. abbreviated AVM

NP 1

synsem

local

category

[head nounsubcat 〈〉

]

content|index 1

S: 1

synsem

local

category

[head verbsubcat 〈〉

]

content 1

VP: 1

synsem

local

category

[head verb

subcat⟨synsem

⟩]

content 1


Theory: Lexicon

Theory: The Lexicon

The basic lexicon is defined by the Word Principle as part of thetheory. It is an implicational statement defining which of theontologically possible words are grammatical:

word → lexical-entry 1 ∨ lexical-entry 2 ∨ . . .

with each of the lexical entries being descriptions, e.g.:

wordphon <laughs>

synsem|loc

cat

head

[verb

vform fin

]

subcat⟨NP[nom] 1 [3rd,sing]

⟩

content

[laugh’

laugher 1

]


Theory: Lexicon

An example lexicon

word →

phon <drinks>

s|l

cat

head

[verb

vform fin

]

subcat⟨NP[nom] 1 [3rd,sing], NP[acc] 2

⟩

cont

drink’

drinker 1

drunken 2

∨

phon <drink>

s|l

cat

head

[verb

vform fin

]

subcat⟨NP[nom] 1 [plur], NP[acc] 2

⟩

cont

drink’

drinker 1

drunken 2


Theory: Lexicon

∨

phon <give>

s|l

cat

head

[verb

vform fin

]

subcat

⟨NP[nom] 1 [plur], NP[acc] 2 ,

PP[to] 3

⟩

cont

give’

giver 1

gift 2

given 3


Theory: Lexicon

∨

phon <to>

s|l

cat

head

[preposition

pform to

]

subcat⟨NP[acc] 1

⟩

cont[index 1

]

∨

phon <think>

s|l

cat

head

[verb

vform fin

]

subcat⟨NP[nom] 1 [plur], S[fin]: 2

⟩

cont

think’

thinker 1

thought 2


Theory: Lexicon

∨

phon <poets>

synsem|loc

cat

[head noun

subcat 〈〉

]

cont

[index

[per third

num plur

]]

∨

phon <wine>

synsem|loc

cat

[head noun

subcat 〈〉

]

cont

[index

[per third

num sing

]]


Introduction

Different kinds of phrasal constructions in HPSG

For phrases, the dtrs feature has a constituent-structure type:

head-comps-struc

head-marker-struchead-dtr phrasemarker-dtr wordcomp-dtrs elist

head-adjunct-struchead-dtr phraseadjunct-dtr phrasecomp-dtrs elist

head-filler-struchead-dtr phrasefiller-dtr phrasecomp-dtrs elist

head-struchead-dtr signcomp-dtrs list(phrase)

coordinate-structureconj-dtrs set(sign)conjunction-dtr word

constituent-structure


Introduction

The structure of a simple phrase

phrasephon <Kim,walks>

synsem

local|cat

head

[verb

vform fin

]

subcat elist

dtrs

head-comp-struc

head-dtr

word

phon <walks>

synsem|loc|cat|head[verb

vform fin

]

comp-dtrs

⟨

phrase

phon <Kim>

synsem|loc|cat|head[noun

case nom

]

⟩


Introduction

Sketch of an example for head-complement structures

phon <Kim>

synsem|loc|cat|head[noun

case nom

]

phon <walks>

synsem|loc|cat|head[verb

vform fin

]

c h

synsem|loc|cat

head

[verb

vform fin

]

subcat 〈〉


Introduction

Sketch of an example for head-complement structures witha ditransitive verb

[phon <John>

synsem 1

]

phon <gives>

synsem|loc|cat

head 4

[verb

vform fin

]

subcat⟨1 , 2 , 3

⟩

[phon <her>

synsem 2

] [phon <the book>

synsem 3

]h c c

[synsem|loc|cat

[head 4

subcat⟨1⟩]]

c h

[synsem|loc|cat

[head 4

subcat 〈〉

]]


Introduction

Head-Feature Principle:

The head value of any headed phrase is structure-shared with thehead value of the head daughter.

[phrasedtrs headed-structure

]→

[synsem|loc|cat|head 1dtrs|head-dtr|synsem|loc|cat|head 1

]


Introduction

Subcategorization Principle:

In a headed phrase (i.e. a phrasal sign whose dtrs value is of sorthead-struc), the subcat value of the head daughter is theconcatenation of the phrase’s subcat list with the list (in order ofincreasing obliqueness) of synsem values of the complementdaughters.


Introduction

Subcat Principle: (cont.)[dtrs headed-structure

]→synsem|loc|cat|subcat 1

dtrs

[head-dtr|synsem|loc|cat|subcat 1 ⊕ 2

comp-dtrs synsem2sign(2)

]

with ⊕ standing for list concatenation, i.e., append, defined asfollows e-list ⊕ 1 := 1 .[

first 1rest 2

]⊕ 3 :=

[first 1rest 2 ⊕ 3

].

and synsem2sign encoding conversion of synsem to sign lists definedas follows

synsem2sign(e-list

):= e-list.

synsem2sign

([first 1rest 2

]):=

[first

[synsem 1

]

rest synsem2sign(2)].


Immediate Dominance Schemata

Immediate Dominance Principle (for English):

[phrase

dtrs headed-struc

]→

synsem|loc|cat

head

([verb

inv −

]∨ ¬ verb

)

subcat 〈〉

dtrs

head-comp-struchead-dtr phrase

comp-dtrs⟨sign

⟩

(Head-Subject)

∨

synsem|loc|cat

head

([verbinv −

]∨ ¬ verb

)

subcat⟨synsem

⟩

dtrs

[head-comp-struchead-dtr word

]

(Head-Complement)



Immediate Dominance Principle (for English):

∨

synsem|loc|cat

head

[verbinv +

]

subcat 〈〉

dtrs

[head-comp-struc

head-dtr word

]

(Head-Subject-Complement)

∨[dtrs

[head-marker-strucmarker-dtr|synsem|loc|cat|head marker

]](Head-Marker)

∨dtrs

head-adjunct-struc

adj-dtr|synsem|loc|cat|head|mod 1

head-dtr|synsem 1

(Head-Adjunct)



Head-Subject Phrases (Schema 1)

[synsem 2

] [synsem|loc|cat

[head 1

subcat⟨2⟩]]

c h

[synsem|loc|cat

[head 1

subcat 〈〉

]]



Head-Complement Phrases (Schema 2)

[synsem|loc|cat

[head 1

subcat⟨2 , 3 ,...,n

⟩]] [

synsem 3].. .

[synsem n

]h c c

[synsem|loc|cat

[head 1

subcat⟨2⟩]]



Head-Subject-Comp Phrases (Schema 3)

[synsem|loc|cat

[head 1

subcat⟨2 ,...,n

⟩]] [

synsem 2].. .

[synsem n

]h c c

[synsem|loc|cat

[head 1

subcat 〈〉

]]



Sketch of an example with an auxiliary

[phon <John>

synsem 1

]

phon <can>

synsem|loc|cat

head 3

verb

vform fin

aux +

inv −

subcat⟨1NP

[nom

], 2VP

[bse

]⟩

[phon <go>

synsem 2

]h c

phon <can, go>

synsem|loc|cat[head 3

subcat⟨1⟩]

c h

phon <John, can, go>


subcat 〈〉

]



Sketch of an example with an inverted auxiliary

phon <can>

synsem|loc|cat

head 3

verb

vform fin

aux +

inv +

subcat⟨1NP

[nom

], 2VP

[bse

]⟩

[phon <John>

synsem 1

] [phon <go>

synsem 2

]h c c

phon <can, John, go>


subcat 〈〉

]



Complementizer as heads

Some verbs select a complementized sentential complementheaded by a verb with a specific verb form:

(9) I demand that he leave/*leaves immediately.



Markers

marker determiner

[functionalspec synsem

]

adjective verb noun preposition . . .

substantive

head

unmarked

that for

comp. . .

marked

marking



Marking Principle:

In a headed structure, the marking value coincides with that ofthe marker daughter if there is one, and with that of the headdaughter otherwise.


]→

synsem|loc|cat|marking 1

dtrs

[head-mark-strucmarker-dtr|synsem|loc|cat|marking 1

]

∨synsem|loc|cat|marking 1

dtrs

[¬head-mark-struchead-dtr|synsem|loc|cat|marking 1

]



Lexical entry of the marker that

phon <that>

synsem—loc—cat

head

mark

spec

loc—cat

head

[verb

vform fin ∨ bse

]

marking unmarked

subcat 〈〉

subcat 〈〉marking that



Sketch of an example for a head-marker structure

phon <that>

synsem|loc|cat

head

[mark

spec 3

]

subcat 〈〉marking 1 that

phon <John laughs>

s 3

loc|cat

head 2

[verb

vform fin

]

marking unmarked

subcat 〈〉

m h

phon <that,John,laughs>

s|loc|cat

head 2

subcat 〈〉marking 1


Determiners

SPEC Principle:

If a nonhead daughter in a headed structure bears a spec value, itis token-identical to the synsem value of the head daughter.

[phrase

dtrs[(marker-dtr ∨ comp-dtrs|first) |synsem|loc|cat|head functional

]]

→

[dtrs

[(marker-dtr ∨ comp-dtrs|first) |synsem|loc|cat|head|spec 1

head-dtr|synsem 1

]]


Determiners

Determiners

Common nouns subcategorize for their determiners:

phon <book>

synsem|loc

cat

head noun

subcat

⟨[loc|cat

[head det

subcat 〈〉

]]⟩

cont

index 1

restr

[reln book

inst 1

]


Determiners

Determiners (cont.)

Determiners select their nominal sister:

phon <every>

synsem|loc

cat

head

det

spec

loc

cat

[head noun

subcat⟨DetP

⟩]

cont 2

subcat 〈〉

cont 5

quant

det forall

restind 2

qstore{5}


Determiners

Sketch of the head-comp phrase every book

phon <every>

synsem 4

loc

cat

head

[det

spec 3

]

subcat 〈〉

cont

[det forall

restind 2

]

phon <book>

s 3

loc

cat

[head 5

[noun

]

subcat⟨4⟩

]

cont 2

c h

phon <every,book>

s|loc

cat

[head 5

subcat 〈〉

]

cont 2


Possessives

Lexical entry for the possessive pronoun my

phon <my>

synsem|loc

cat

head

det

spec N’:

[index 1

restr 2

]

subcat 〈〉

cont|index 3

ref

per 1st

num sing

context|c-indc|speaker 3

qstore

det the

restind

index 1

restr

reln poss

possessor 3

possessed 1

∪ 2


Possessives

Lexical entry for possessive ’s

phon <’s>

synsem|loc

cat

head

det

spec N’:

[index 1

restr 2

]

subcat

⟨NP: 5

[npro

index 3

]⟩

cont 5

qstore

det the

restind

index 1

restr

reln poss

possessor 3

possessed 1

∪ 2


Possessives

Sketch of an example for a determiner phrase

phon <Mary’s>

synsem|loc

cat

head

det

spec N’:

[index 1

restr 2

]

subcat 〈〉

cont

[npro

index 3

]

context|backgr

reln naming

bearer 3

name mary

qstore

det the

restind

index 1

restr

reln poss

possessor 3

possessed 1

∪ 2


Adjuncts

Lexical entry of an attributive adjective

wordphon <red>

synsem|loc

cat|head

adj

prd −

mod|loc

cat

[head noun

subcat⟨[loc|cat|head det

]⟩]

content

[index 1 index

restr 2

]

cont

index 1

restr

{[rel red

arg1 1

]}∪ 2


Adjuncts

Semantics Principle

In a headed phrase, the content value is token-identical to thatof the adjunct daughter if the dtrs value is of sort head-adj-struc,and with that of the head daughter otherwise.


]→

synsem|loc|content 1

dtrs

[head-adj-strucadjunct-dtr|synsem|loc|content 1

]

∨synsem|loc|content 1

dtrs

[¬head-adj-struchead-dtr|synsem|loc|content 1

]


Adjuncts

Sketch of an example for a head-adjunct structure

phon <red>

s|loc

cat|head

adj

prd −mod 3

cont 4

phon <book>

s 3

[loc|cat

[head 2 noun

subcat⟨1[loc|cat|head det

]⟩]]

a h

phon <red, book>

s|loc

cat

[head 2

subcat⟨1⟩]

cont 4


Raising/EquiReview of ID Schemata

[phrasedtrs headed-struc

]→

synsem|loc|cat

head

([verbinv −

]∨ ¬ verb

)

subcat 〈〉

dtrs

head-comps-struchead-dtr phrase

comp-dtrs⟨sign

⟩

(Hd-Sbj)

∨

synsem|loc|cat

head

([verbinv −

]∨ ¬ verb

)

subcat⟨synsem

⟩

dtrs

[head-comps-struchead-dtr word

]

(Hd-Cmp)

∨

synsem|loc|cat

head

[verbinv +

]

subcat 〈〉

dtrs

[head-comps-struchead-dtr word

]

(Hd-Sbj-Cmp)


Towards an analysis: Unsaturated Complements

Where does it say in these ID schemata that every subcategorizeditem must be realized?

In English, many verbs and adjectives subcategorize for anunsaturated complement.

In other words, a complement can be specified as[subcat

⟨NP

⟩], rather than

[subcat 〈〉

]

The Head-Subject Schema allows for this.And this will give consider access to the lower subject, as wellas its own subject.


Subject oriented raising verbs

wordphon <seem>

synsem

synsem

local

local

cat

cat

head

[verb

vform bse

]

subcat⟨1 , VP

[inf, subcat

⟨1⟩]: 2

⟩

cont

[seem

soa-arg 2

]


Subject oriented equi verbs

wordphon <try>

synsem

synsem

local

local

cat

cat

head

[verb

vform bse

]

subcat⟨NP 1 , VP

[inf, subcat

⟨NP 1

⟩]: 2

⟩

cont

try

tryer 1 ref

soa-arg 2


Object oriented raising verbs

wordphon <believe>

synsem

synsem

local

local

cat

cat

head

[verb

vform bse

]

subcat⟨NP 1 , 2 , VP

[inf, subcat

⟨2⟩]: 3

⟩

cont

believe

believer 1 ref

soa-arg 3


Object oriented equi verbs

wordphon <persuade>

synsem

synsem

local

local

cat

cat

head

[verb

vform bse

]

subcat⟨NP 1 , NP 2 , VP

[inf, subcat

⟨NP 2

⟩]: 3

⟩

cont

persuade

persuader 1 ref

persuadee 2 ref

soa-arg 3


Analysis of: It seems to rain

wordp <it>

s 5

l

cat

[head noun

subc 〈〉

]

cont[index it

]

word

p <seems>

s|l

cat

head 9

[verb

vform fin

]

subc⟨5 , 8VP

[inf, subc

⟨5⟩]: 7

⟩

cont 10

[seem

soa-arg 7

]

phrase

p <to rain>

s 8

l

cat

head 6

[verb

vform inf

]

subc⟨5NP it

⟩

cont 7 rain

h c

phrasep <seems to rain>

s|l

cat

[head 9

subc⟨5⟩]

cont 10

c h

phrase

p <It seems to rain>

s|l

cat

[head 9

subc 〈〉

]

cont 10


Analysis of: *It wants to rain

word

p <wants>

s|l

cat

head 9

[verb

vform fin

]

subc⟨3NP 11 , 8VP

[inf, subc

⟨5NP 11

⟩]: 7

⟩

con 10

want

wanter 11 ref

soa-arg 7

phrase

p <to rain>

s 8

l

cat

head 6

[verb

vform inf

]

subc⟨5NP it

⟩

con 7 rain

h c

phrase

p <wants to rain>

s |l

cat

[head 9

subc⟨3⟩]

con 10


Unbounded dependency constructions (UDCs)

To account for UDCs, we will use the feature slash.

Named after the categorial grammar categories (S/NP as an Smissing an NP).

This is a non-local feature which:

originates with a trace,gets passed up the tree,and is finally bound by a filler

An example for a strong UDC

Johni

NP

we

NP

know

V

she

1NP

likes

V[loc|cat|subcat

⟨1 , 2

⟩]

i

2NPh c

VP[loc|cat|subcat

⟨1⟩]

c h

S[loc|cat|subcat 〈〉

]h c

VPc h

Sf h

S


The bottom of a UDC: Traces

wordphon 〈〉

synsem

local 1

nonloc

[inherited|slash

{1}

to-bind|slash {}

]

phonologically null, but structure-shares local and slashwe’ll talk about to-bind later


Traces

Because the local value of a trace is structure-shared with theslash value, constraints on the trace will be constraints on thefiller.

For example, hates specifies that its object be accusative, andthis case information is localSo, the trace has

[synsem|local|cat|head|case acc

]as part

of its entry, and thus the filler will also have to be accusative

(10) *Hei/Him i , John likes i


The middle of a UDCThe Nonlocal Feature Principle (NFP)

For each nonlocal feature, the inherited value on the mother isthe union of the inherited values on the daughter minus theto-bind value on the head daughter.

In other words, the slash information (which is part ofinherited) percolates “up” the tree.

This allows the all the local information of a trace to “moveup” to the filler.


The top of a UDC: Filler-head structuresFiller-head schema

[phrasedtrs head-filler-struc

]→

head-dtr|synsem

loc|cat

head

[verb

vform fin

]

subcat 〈〉

nonloc

[inherited|slash element

(1)

to-bind|slash{1}

]

filler-dtr|synsem|local 1


The top of a UDC: Filler-head structuresExplanation of the schema

Filler and trace are identified as the exact same thing (as faras their local structure is concerned)The trace is “bound” by the to-bind feature; this preventsthe slash value from going any higher in the treeOnly saturated finite verbs (i.e., sentences) license suchstructures


The top of a UDC: Filler-head structuresExample for a structure licensed by the filler-head schema

[local 1

] [nloc

[inherited|slash

{. . . , 1 ,. . .

}

to-bind|slash{1}

]]f h

[nloc|inherited|slash {}

]


The analysis of the strong UDC example

Johni

NP[local 3

]

we

NP

know

V

she

1NP

likes

V[loc|cat|subcat

⟨1 , 2

⟩

nonloc|to-bind|slash {}

]

i

NP

2

[loc 3

nloc|inher|slash{

3}]

h c

VPloc|cat|subcat

⟨1⟩

nloc

[inherited|slash

{3}

to-bind|slash {}

]

c h

Sloc|cat|subcat 〈〉nloc

[inherited|slash

{3}

to-bind|slash {}

]

h c

VP[nloc

[inherited|slash

{3}

to-bind|slash {}

]]c h

S[nloc

[inherited|slash

{3}

to-bind|slash{

3}]]

f h

S[nloc|inherited|slash {}

]

Documents

Head-Driven Phrase Structure Grammar (HPSG)cl.indiana.edu/~md7/15/614/slides/06-hpsg/06-hpsg-1x3.pdfIntroduction Signs Phrases Raising/Equi UDCs Head-Driven Phrase Structure Grammar