What levels of linguistic rep resentation determine o r - - PDF document

what levels of linguistic rep resentation determine o r
SMART_READER_LITE
LIVE PREVIEW

What levels of linguistic rep resentation determine o r - - PDF document

What levels of linguistic rep resentation determine o r constrain the semantic level? W e b elieve that the f-structure is the p rima ry level that con- strains semantic interp retation. Of course, info


slide-1
SLIDE 1 Do w e need a linguistic rep resentation
  • f
semantic info rmation?
  • Y
es, fo r purp
  • ses
  • f
talking ab
  • ut
semantics.
  • Is
the rep resentation \disp ensible"? P erhaps { dep ends
  • n
choice
  • f
semantic theo ry . Semantic structures can b e re- lated to a level
  • f
rep resentation
  • f
meaning (Discourse Rep- resentation Structures, Situation-Theo retic Infons, fo rmulas
  • f
intensional logic)
  • r
directly to a mo del. 20 What levels
  • f
linguistic rep resentation determine
  • r
constrain the semantic level?
  • W
e b elieve that the f-structure is the p rima ry level that con- strains semantic interp retation. Of course, info rmation from
  • ther
levels, such as c-structure, ma y also b e relevant. The relation b et w een the semantic structure and these
  • ther
levels ma y b e enco ded directly b y a p rojection function,
  • r
indirectly as a comp
  • sition
  • f
p ro- jection functions b et w een
  • ther
levels.
  • Even
if
  • ther
levels (e.g. c-structure) constrain scop e, w e needn't have a level
  • f
rep resentation at which info rmation from b
  • th
levels is enco ded. W e can talk ab
  • ut
the relation b et w een t w
  • levels
in addition to relations within a level. 19
slide-2
SLIDE 2 What kind
  • f
info rmation needs to b e rep resented in a gramma r?
  • F-structure:
rep resents syntactic a rgument structure
  • Semantic
structure: rep resents semantic t yp e structure with no syntactic a rgument structure reex
  • Glue
language: constrains ho w linguistic structures deter- mine the assembly
  • f
meanings (issues ab
  • ut
relative scop e, scop e islands, t yp e raising, . . . ) { \grammatical semantics"
  • Meaning
language: Meanings 18 Building up a quantier: \Every p erson" g : 2 4 SPEC `every' PRED `p erson' 3 5
  • g
  • :
" V AR [ ] RESTR [ ] # every : 8R ; P ; S: [8Y :(g
  • V
AR) ; Y
  • (g
  • RESTR
) ; R (Y )] [8X :g
  • ;
X
  • S
; P (X )]
  • S
; ev er y (per son; P ) p erson : 8X :(g
  • V
AR) ; X
  • (g
  • RESTR
) ; per son(X ) every p erson : 8P ; S:[8X :g
  • ;
X
  • S
; P (X )]
  • S
; ev er y (per son; P ) 17
slide-3
SLIDE 3 Quantication: \Every
  • ne
w alks." f
  • :[
]
  • f
: 2 6 4 PRED `w alk<(" SUBJ)>' SUBJ g : h PRED `every
  • ne'
i 3 7 5
  • g
  • :[
] every
  • ne
: 8P ; S:[8X :g
  • ;
X
  • S
; P (X )]
  • S
; ev er y (per son; P ) w alks : 8X :(f SUBJ)
  • ;
X
  • f
  • ;
w al k (X ) every
  • ne
w alks : f
  • ;
ev er y (per son; w al k ) 16 \John w alks." f
  • :[
]
  • f
: 2 4 PRED `w alk<(" SUBJ)>' SUBJ g :
  • PRED
`John'
  • 3
5
  • g
  • :[
] g
  • ;
j
  • hn
8X :(f SUBJ )
  • ;
X
  • f
  • ;
w al k (X ) f
  • ;
w al k (j
  • hn)
15
slide-4
SLIDE 4 \John" g :
  • PRED
`John'
  • g
  • :[
] ; j
  • hn
14 An a rchitecture fo r the syntax-semantics interface: Assemble meanings with instructions in a logical language
  • Use
\glue language", linea r logic, to sp ecify ho w to put meanings together
  • Meaning
language: y
  • ur
choice; w e use higher-o rder inten- sional logic 13
slide-5
SLIDE 5 Problems with function application: 2 6 6 6 6 6 4 REL
  • V
AR [ ] RESTR " REL P ARG [ ] # 3 7 7 7 7 7 5 h ARG [ ] i ) " REL P ARG [ ] # X :P (X ) (Y ) ) P (Y ) 12 Problems with va riable binding: 2 6 4 [ ] h [ ] i 3 7 5 2 6 6 6 6 6 6 4 2 6 6 6 6 6 4 REL 8 V AR [ ] RESTR " REL P ARG [ ] # 3 7 7 7 7 7 5 3 7 7 7 7 7 7 5 2 6 6 6 6 6 6 6 6 6 4 [ ] 2 6 6 6 6 6 4 REL 8 V AR [ ] RESTR " REL P ARG [ ] # 3 7 7 7 7 7 5 3 7 7 7 7 7 7 7 7 7 5 11
slide-6
SLIDE 6 Commonalities:
  • Sepa
rate rep resentation
  • f
syntactic and semantic info rma- tion
  • F
  • rm
  • f
meaning: attribute-value structure; gives (mo re
  • r
less) undersp ecied rep resentation
  • f
semantic info rmation
  • Meaning
related directly to c-structure
  • r
f-structure
  • Meaning
assembly b y analysis
  • f
f-structure
  • r
accumulation
  • f
constraints 10 Halvo rsen and Kaplan (1988), Projections and semantic description in LF G:
  • F
  • rm
  • f
meaning: attribute-value structure
  • Meaning
determined b y p rojection from c-structure, indi- rectly related to f-structure
  • Meaning
assembled b y accumulation
  • f
constraints
  • n
attribute- value pairs 9
slide-7
SLIDE 7 Halvo rsen and Kaplan (1988), Projections and semantic description in LF G:
  • 2
4 PRED `w alk<(" SUBJ)>' SUBJ
  • PRED
`John'
  • 3
5 S NP VP N V John w alks " PRED
  • REL
w alk
  • ARG1
John # 8 F enstad et al. (1987), Situations, Language, and Logic:
  • F
  • rm
  • f
meaning: attribute-value structure, the sitschema, rep resenting a fo rmula in Situation Semantics
  • Meaning
determined (in p rinciple) b y phonology , mo rphology , syntax, context
  • Meaning
assembled b y accumulation
  • f
constraints
  • n
attribute- value pairs 7
slide-8
SLIDE 8 F enstad et al. (1987), Situations, Language, and Logic: 2 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 4 FSTRUCT 2 4 PRED `w alk<(" SUBJ)>' SUBJ
  • PRED
`John'
  • 3
5 SITSCHEMA 2 6 6 6 6 6 6 6 6 6 6 6 6 6 4 REL w al k ARG1 h IND j
  • hn
i LOC 2 6 6 6 6 6 4 IND I N D :1 COND 2 6 4 REL
  • ARG.1
[ ] 1 ARG.2 l d 3 7 5 3 7 7 7 7 7 5 3 7 7 7 7 7 7 7 7 7 7 7 7 7 5 3 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 5 6 Halvo rsen (1983), Semantics fo r LF G:
  • F
  • rm
  • f
meaning rep resentation: attribute-value structure { F-structure and fo rmula
  • f
intensional logic a re disp ensible { Dierent meaning language is p
  • ssible
  • Meaning
determined b y f-structure
  • Meaning
assembled b y analysis
  • f
f-structure 5
slide-9
SLIDE 9 \John w alks." f k = 2 4 PRED `w alk<(" SUBJ)>' SUBJ
  • PRED
`John'
  • 3
5 M k = 2 6 6 6 6 6 4 PREDICA TE w al k ARG1 2 6 6 4 CM P :P (j
  • hn)
MODE CM PM P :P (j
  • hn)
3 7 7 5 3 7 7 7 7 7 5 F
  • rmula
  • f
intensional logic: w al k
  • (j
  • hn)
4 Halvo rsen (1983), Semantics fo r LF G: PRED-ARG conguration If f k is an f-structure
  • f
the fo rm 2 4 s 1 v 1 . . . s n v n 3 5 containing some v i that has an a rgument list, then (M k PREDICA TE) = M s i and fo r < j
  • m,
(M k ARGj ) = M l where m is the numb er
  • f
thematic a rguments
  • f
the semantic fo rm in s i , and M l is the semantic structure asso ciated with the f-structure designato r in the jth a rgument p
  • sition.
3
slide-10
SLIDE 10
  • What
is the fo rm
  • f
the meaning rep resentation?
  • What
is the relation
  • f
the meaning rep resentation to
  • ther
levels?
  • Ho
w a re meanings put together? 2 Levels
  • f
semantic rep resentation in LF G Ma ry Dalrymple, John Lamping, and Vija y Sa rasw at Semantics W
  • rkshop
at the LF G Collo quium and W
  • rkshops
Grenoble, F rance August 26, 1996 1