semantic composition - stanford university · (tv) given a syntactic structure vp v pn, jvpk=...
TRANSCRIPT
Semantic composition
Chris Potts, Ling 130a/230a: Introduction to semantics and pragmatics, Winter 2020
Jan 28
1 Overview
• This handout describes our theory of semantic composition. It’s a bare-bones theory, but still
powerful.
• The most important conceptual move is to interpret lexical items as functions. Meanings
might not literally be functions, but, following Lewis’s advice, functions at least do what mean-
ings do.
• The grammar is defined by a set of rules for interpreting syntactic structures. Some of these
rules come from the optional Heim & Kratzer reading, while others are new. My notation is
slightly different (to more strongly suggest a computational perspective), but the underlying
ideas are the same as in H&K. (I’ve set up these correspondences largely to try to build some
bridges into this advanced optional reading.)
• We need a bunch of rules in order to respect the syntactic structures. However, all the rules
just do function application.
• So, in essence, if you know the lexical meanings of your language and you can put them
together according to the syntactic rules, then the only other concept you need to be a full-
fledged interpreter is function application.
2 Basic semantic objects
2.1 Truth values
I use T for truth and F for falsity. (Heim & Kratzer use 1 and 0, respectively.)
2.2 Universe
U =⇢
, , ,
�
3 Characteristic sets and characteristic functions
Characteristic function of a set For a universe of entities U and any set A✓ U , the characteristic
function for A is the function that maps every member of A to T and all members of (U � A) to F.
Characteristic set of a function If f is a function into truth values, the characteristic set of f is
{x | f (x) = T}
Set Function
⇢, ,
�
26666666666664
7! T
7! T
7! F
7! T
37777777777775
⇢,
�
26666666666664
7!
7!
7!
7!
37777777777775
;
26666666666664
7!
7!
7!
7!
37777777777775
Set Function
26666666666664
7! T
7! F
7! F
7! F
37777777777775
26666666666664
7! T
7! T
7! T
7! T
37777777777775
26666666666664
7! F
7! F
7! F
7! F
37777777777775
Emas
amassed
Maggie
T
F UF
T
F
F
FF
O t GET
sizes 3IE
4 Notation for describing functions
4.1 Functions
26666666666664
7! T
7! T
7! T
7! F
37777777777775
CHILD(x)
1 if x 2ß
, ,
™
2 return T3 else return F
�x
0B@T if x 2ß
, ,
™, else F
1CA
4.2 Function application
0B@�x
0B@T if x 2ß
, ,
™, else F
1CA
1CAÅ ã
=
T if 2ß
, ,
™, else F =
T
5 Semantic lexicon
5.1 PNs
JMaggieK= JLisaK= JBartK= JHomerK=
5.2 Ns
(1) JSimpsonK= �x
0B@T if x 2ß
, , ,
™, else F
1CA
(2) JchildK= �x
0B@T if x 2ß
, ,
™, else F
1CA
type'ifunctionfronentitiattrutludes
CHILD
i
To
zeuioyfch.ldDCes.frouztahfowonta.es
1
the.EE4 Tifxe elseFµremind sore.se
Characteristicfunctionoftheses X
Type: entities
Type: functions from entities to truth values
(3) JstudentK= �x
0B@T if x 2ß
,
™, else F
1CA
(4) JparentK= �x
0B@T if x 2ß ™
, else F
1CA
5.3 Intransitive Vs
(5) JskateboardsK= �x
0B@T if x 2ß
,
™, else F
1CA
(6) JstudiesK= �x
0B@T if x 2ß ™
, else F
1CA
(7) JintrospectsK= �x
0B@T if x 2ß
,
™, else F
1CA
(8) JspeaksK=
5.4 Transitive Vs
(9) JteaseK= �y
0BBBBBBBBBBB@
�x
0BBBBBBBBBBB@
T if hx , yi 2
8>>>>>>>>>><>>>>>>>>>>:
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑
9>>>>>>>>>>=>>>>>>>>>>;
, else F
1CCCCCCCCCCCA
1CCCCCCCCCCCA
Isametypeandmeaning
framework
as Ns
t1x Tif EU B elseF
I
iA teased Labar
tease PNE µypekassiaieasess37D hs.DGb
directors't E a
Type: functions from entities to truth values
Type: functions from entities into functions from entities to truth values
(10) JadmiresK= �y
0BBBBB@�x
0BBBBB@T if hx , yi 2
8>>>>><>>>>>:
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑
9>>>>>=>>>>>;
, else F
1CCCCCA
1CCCCCA
(11) JlovesK=
5.5 Adjectives
(12) JfemaleK= � f
0B@�x
0B@T if x 2ß
,
™and f (x) = T, else F
1CA
1CA
(13) JmaleK= � f
0B@�x
0B@T if x 2ß
,
™and f (x) = T, else F
1CA
1CA
(14) JhungryK= � f
0B@�x
0B@T if x 2ß
,
™and f (x) = T, else F
1CA
1CA
(15) JallegedK= � f (�x (T if someone claimed that f (x) = T, else F))
(16) JSpringfieldianK=
Setd ordered
fqfIIIIIE.es GxGitLx.DEPienes eluFD
r tfemaleskateboarder
l
framework 7f Xx 3gtie afootie function
GlendeDCashdeboarderD
XfGxTis xE Ehsaan.gsauefkdT.elseFDEshdeboa.eeiD
XxCTifxCElissaErassied3anaffshdeboarderDGd I elseF
WTifxE asadAnansie andGYGifyefabartkdhonerH.ieseF x elseFD
Gif xEElisaGrossed aneTifxEEbartD.Ehonerd3eeseFelseF
cacti xcLEhsaa.serao.ie are C aboard.GhorerB elseF
Type: functions from functions from entities to truth values into functions from entities to truth values
Intersective: T if the argument is in the first set and true of the characteristic function of the second
5.6 Negation
To be completed:
(17) � f
✓�x
✓ ◆◆
5.7 Quantificational determiners
(18) JeveryK= � f
�g
ÄT if {x | f (x) = T} ✓ {x | g(x) = T} , else F
äã
(19) JsomeK= � f
�g
ÄT if {x | f (x) = T}\ {x | g(x) = T} 6= ;, else F
äã
(20) JnoK= � f
�g
ÄT if {x | f (x) = T}\ {x | g(x) = T}= ;, else F
äã
(21) JmostK=
(22) JthreeK=
(23) JfewK=
Etta LEIF Bartnotshieboards
amodelforhowto
approach
thatquestion
onthe
Tiff Felse XfGxEndalf assignment
µgJA B
X pi ToutputoutputjnputddIFH.IE
tfGgGif t I IIcharset I 2 etat
charselff x fCx T
XfXg Tif IchorsetCArcharsetts lZ3elseFD
Type: functions from functions from entities to truth values into (functions from entities to truth values into truth values)
6 Semantic grammar
(NB) Given a syntactic structure X
Y
, JXK= JYK
(S) Given a syntactic structure S
PN VP
, JSK= JVPK(JPNK)
(A) Given a syntactic structure NPj
AP NPi
, JNPjK= JAPK(JNP
iK)
(N) Given a syntactic structure VPj
not VPi
, JVPjK= JnotK(JVP
iK)
(With apologies to the morphosyntax of English for providing no do support.)
(TV) Given a syntactic structure VP
V PN
, JVPK= JVK(JPNK)
(Q1) Given a syntactic structure QP
D NP
, JQPK= JDK(JNPK)
(Q2) Given a syntactic structure S
QP VP
, JSK= JQPK(JVPK)
metavandices
overthesyntacticcategories
PNlPNBlade
where CHD is
A wellformed the
functorsure't
reverse Days
nee is
i am j justhelpus distinguish thetwoidenticalcategorylabels
Quantifier
b
TDeterminer
7 Illustrations
(24) S
PN
Bart
VP
V
skateboards
�x
0BBB@T if x 2
ß,
™, else F
1CCCA
✓ ◆=
T if 2ß
,
™, else F
�x
0BBB@T if x 2
ß,
™, else F
1CCCA
JskateboardsK
JskateboardsK
(25) NP
AP
A
female
NP
N
student
� f
0BBBBB@�x
0BBBBB@T if x 2
ß,
™and f (x) = T, else F
1CCCCCA
1CCCCCA
0BBBBB@�x
0BBBBB@T if x 2
ß,
™, else F
1CCCCCA
1CCCCCA
� f
0BBBBB@�x
0BBBBB@T if x 2
ß,
™and f (x) = T, else F
1CCCCCA
1CCCCCA
JfemaleK
�x
0BBBBB@T if x 2
ß,
™, else F
1CCCCCA
JstudentK
S
NBis
ANB
NB
NB NB Thismeaningiscorrectbut
wouldnotsatisfythe requirement
of thehomeworkthatallsubstitutions beperformed
(26) S
PN
Maggie
VP
V
teases
PN
Bart
T if h , i 2
8>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>:
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑
9>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>;
, else F
�x
0BBBBBBBBBBBBBBBBB@
T if hx , i 2
8>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>:
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑
9>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>;
, else F
1CCCCCCCCCCCCCCCCCA
�y
0BBBBBBBBBBBBBBBBB@
�x
0BBBBBBBBBBBBBBBBB@
T if hx , yi 2
8>>>>>>>>>>>>>>>><>>>>>>>>>>>>>>>>:
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑,
≠,
∑
9>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>;
, else F
1CCCCCCCCCCCCCCCCCA
1CCCCCCCCCCCCCCCCCA
JteaseK
S
µ
(27) S
QP
D
every
NP
N
child
VP
V
skateboards
T if
�x | JchildK(x) = T
✓�
x | JskateboardsK(x) = T
, else F
�g
ÄT if
�x | JchildK(x) = T
✓ {x | g(x) = T} , else F
ä
� f
�g
ÄT if {x | f (x) = T} ✓ {x | g(x) = T} , else F
äã
JeveryK
JchildK
JchildK
JchildK
JskateboardsK
JskateboardsK
JskateboardsK
QP2
QPl NB
NB NB NB
NB
(28) S
PN
Homer
VP
V
teases
PN
Lisa
(29) S
PN
Homer
VP
not VP
V
teases
PN
Lisa
(30) QP
D
a
NP
Simpson
S
N4Elisac Laid
akaesb3deF
GfGzL ftp.DJGYG Lg.EusuDEKaDakoesb3elseFDGzEE3EJGgGit4ausakkasiateamb3elur.DK
Gcc Giftisabella.ir.aleausB.elser
(31) S
QP
D
no
NP
N
parent
VP
V
skateboards
(32) S
QP
D
every
NP
AP
A
female
NP
N
child
VP
V
studies
(33) S
QP
D
a
NP
AP
A
hungry
NP
N
child
VP
not VP
V
studies