Lexical-Functional Grammar (LFG) Motivation for LFG LFG in a ...

The resulting f-structure for the example sentence
Part IV: A few notes
2 2 3
3
spec a
6 7
subj f 2: 4num
sg 5
pred ’girl’
tense past
pred ‘hand ’
2
3
f 1, f 3:
spec the
6
7
obj f 4: 4num
sg 5
pred ‘baby’
2 3
spec a
6 6 7
4obj2 f 5 : 4num
sg
7
5
5
pred ‘toy’
We’ll finish up the unit on LFG by looking quickly at:
• Extraction
• The syntax-semantics interface
• Computational issues
43/50
44/50
Extraction: Functional uncertainty
Extraction example
The way extraction is handled in LFG is by functional uncertainty: a functional
equation sets up a relation between some initial, extracted object with a
grammatical function (GF) later in the sentence.
Which GF is left unspecified, e.g.:
(29) CP → XP
(↑focus) = ↓
(↑focus) = (↑comp* gf)
C’
↑ = ↓
This says that the focus element is equated with some gf after a path of comp
values
(30) What do you think Chris bought?
2 h
i
3
focus pred ’what’
pred ’think’
h
i
subj pred ’pro’
(31)
2
3
pred ’buy’
h
i
6comp
6
4 4
subj pred ’Chris’
7
5
obj
The principle of completeness ensures that bought has a realized object, and the
functional equation fills it in.
45/50
46/50
Extraction example 2
The syntax-semantics interface
(32) What do you think Chris hoped David bought?
2 h
i
3
focus pred ’what’
pred ’think’
h
i
subj pred ’pro’
2
3
pred ’hope’
(33)
h
i
subj pred ’Chris’
2
3
comp
pred ’buy’
h
i
6 6comp
6
4 4 4subj
pred ’David’
7
5
obj
F-structures are fairly closely linked to the semantics of a sentence, but the
details still have to be worked out:
One theory of the syntax-semantics interface is glue semantics
• A compositional form of semantics, which maps from the f-structure to a
semantic formula
– The σ function performs the mapping
• Employs linear logic
– Each premise (i.e., word) must be used once and can only be used once
– It is thus resource-driven, requiring every word to contribute to the meaning
47/50
48/50