Upload
mattox
View
32
Download
0
Tags:
Embed Size (px)
DESCRIPTION
Binding Theory in LTAG. Lucas Champollion University of Pennsylvania [email protected]. Overview. Binding Theory (BT) and its local domains Previous work: Condition A This proposal: Conditions A, B, C Discussion. Binding theory: A reminder. - PowerPoint PPT Presentation
Citation preview
June 7th, 2008 TAG+9 1
Binding Theory in LTAG
Lucas ChampollionUniversity of Pennsylvania
June 7th, 2008 TAG+9 2
Overview
Binding Theory (BT) and its local domains Previous work: Condition A This proposal: Conditions A, B, C Discussion
June 7th, 2008 TAG+9 3
Binding theory: A reminder Condition A: reflexives must be locally bound
Johnj thinks [ Billb likes himself*j / b / *[other] ] Condition B: pronouns must be locally free
Johnj thinks [ Billb likes himj / *b / [other] ] Condition C: full noun phrases must be free
*[ Johnj likes Johnj ] *Johnj thinks [ Mary likes Johnj ]
June 7th, 2008 TAG+9 4
Binding theory in LTAG
LTAG’s local domain = the verbal elementary tree and its arguments (but not its adjuncts)
Insight from previous work: LTAG and BT have similar local domains
This presentation’s central point: Too many mismatches between local domains We can’t reuse LTAG’s local domain for binding!
June 7th, 2008 TAG+9 5
Previous work reused LTAG’s local domain
NP
loves
V NP
VP
S
himself
John
NP
V S*
VP
S
thinks
he
Condition A
June 7th, 2008 TAG+9 6
Previous work reused LTAG’s local domain
NP
loves
V NP
VP
S
himself
John
NP
V S*
VP
S
thinks
he
Condition A
June 7th, 2008 TAG+9 7
Previous work reused LTAG’s local domain
NP
loves
V NP
VP
S
himself
John
NP
V S*
VP
S
thinks
he
Condition A
June 7th, 2008 TAG+9 8
Previous work reused LTAG’s local domain
NP
loves
V NP
VP
S
him
John
NP
V S*
VP
S
thinks
he
Condition B
June 7th, 2008 TAG+9 9
Ryant and Scheffler (2006) Only Condition A MCTAG set with a
degenerate NP tree Tree-local MCTAG with
flexible composition makes sure that antecedent and reflexive substitute into the same tree
NP
loves
V NP
VP
S
himself
NPi
John
NP
NP*i{ }
June 7th, 2008 TAG+9 10
Kallmeyer and Romero (2007) Only Condition A MCTAG set with a
degenerate VP tree Tree-local MCTAG with
flexible composition makes sure that antecedent and reflexive substitute into the same tree
NP
loves
V NP
VP
S
himself
NPi
John
NP
VP*i{ }(some features omitted)
June 7th, 2008 TAG+9 11
Kallmeyer and Romero’s claim
“Tree-local MCTAG display exactly the extended domain
of locality needed to account for the locality of anaphora binding in a natural way.”-- Kallmeyer and Romero (2007)
June 7th, 2008 TAG+9 12
A counterexample
VP*
opposite
P NP
PP
VP
himself
John
NP
V NP
VP
S
imagined
Cannot be handled by Kallmeyer and Romero (2007) except by flexible composition (which they try to avoid)
Bill
June 7th, 2008 TAG+9 13
ECM: another mismatch of localities
NP
to love
V NP
VP
S
Bill
John
NP
V S*
VP
S
expects
him
Can be handled with an extra feature No lexical ambiguity needed (unlike R&S 2006)
June 7th, 2008 TAG+9 14
Mismatches within Binding Theory
Bill
NP
NP
himself
NP
V NP
VP
S
found
John
NP
NP*
’s
Det NP
N’
NP
picture
N
P NP*
PP
NP
of
Bill
NP
NP
him
NP
V NP
VP
S
found
John
NP
NP*
’s
Det NP
N’
NP
picture
N
P NP*
PP
NP
of
Judgments tested experimentally (Keller and Asudeh ‘01; Runner ‘03)
A B
June 7th, 2008 TAG+9 15
Mismatches within Binding Theory
VP*
near
P
PP
VP
himselfJohn
NP
V NP
VP
S
saw a snake
NP
VP*
near
P
PP
VP
himJohn
NP
V NP
VP
S
saw a snake
NP
A
B
June 7th, 2008 TAG+9 16
How to encode the other conditions? Condition A roughly corresponds to tree-
locality Condition B = “enforced non-locality”? Condition C = ???
Need to propagate an unbounded number of potential antecedents
June 7th, 2008 TAG+9 17
This account in a nutshell Every NP receives three items from its
environment: a list “A” of local potential antecedents a list “B” of local potential antecedents a list “C” of nonlocal potential antecedents
Every NP supplies its own individual variable to its environment
The rest of the grammar is responsible for providing the correct lists to the NP substitution slots
June 7th, 2008 TAG+9 18
Technical innovation: List-valued features
June 7th, 2008 TAG+9 19
Elementary tree for “himself”(Condition A, simplified)
“A reflexive must be locally bound.”
June 7th, 2008 TAG+9 20
Elementary tree for “he”(Condition B)
“A pronoun must be locally free.”
June 7th, 2008 TAG+9 21
Elementary tree for “John” (Condition C)
“A full noun phrase must be free.”
June 7th, 2008 TAG+9 22
Sample derivation
June 7th, 2008 TAG+9 23
Sample derivation
June 7th, 2008 TAG+9 24
Sample derivation
June 7th, 2008 TAG+9 25
Sample derivation
June 7th, 2008 TAG+9 26
Condition C: the default case
Before...
June 7th, 2008 TAG+9 27
Condition C: the default case
...and after unification of top/bottom features
June 7th, 2008 TAG+9 28
Condition C across clauses
Before putting the trees together...
June 7th, 2008 TAG+9 29
Condition C across clauses
The higher tree passesits subject down, then...
June 7th, 2008 TAG+9 30
Condition C across clauses
...unification at the root node propagates the empty list
June 7th, 2008 TAG+9 31
Improvements over previous accounts...
June 7th, 2008 TAG+9 32
Binding into adjuncts
Just propagate everything!
June 7th, 2008 TAG+9 33
Mismatches between domains easily encoded Non-complementary binding conditions easily
handled with separate A and B list features No ad hoc trees needed for picture NPs (unlike
K&R ‘07)
June 7th, 2008 TAG+9 34
C-command violations easily encoded
No need for separate lexical entry Just extrapose subject NP along with its feature structure
(he)
(Himself)
e.g. extraposition: “Himselfi, hei likes.”
June 7th, 2008 TAG+9 35
Improvements at a glance
All conditions are implemented Higher empirical accuracy No lexical ambiguity No flexible composition (K&R 2007) No syntactically unmotivated degenerate
trees (Kallmeyer and Romero, 2008) Better integration with anaphora resolution
(Branco, 2002) No explicit representation of c-command
June 7th, 2008 TAG+9 36
Issues / Future work
Unknown complexity of list-valued features Just a decoration on the trees though -- they do
not rule out any sentences Lack of predictive power
How do we constrain possible feature values? Metagrammar?
Does TAG offer any insights into BT at all?
June 7th, 2008 TAG+9 38
Previous accounts do not interface well with anaphora resolution modules Previous accounts: parser delivers a forest of
indexed trees Johni introduced Billk to himselfi vs.
Johni introduced Billk to himselfk
Problem: Anaphora resolution modules are not prepared to compare entire trees (Branco, 2002)
Our solution outputs a compact set of constraints Following Branco (2002)
June 7th, 2008 TAG+9 39
The grammar of picture NPs
NP*
’s
Det NP
N’
NP
Bill
NP
NP
himself
NP
V NP
VP
S
found
picture
N
P NP*
PP
NP
of
John
NP
June 7th, 2008 TAG+9 40
Missing link problem
NP*
’s
Det NP
N’
NP
Bill
NP
NP
himself
NP
V NP
VP
S
found
picture
N
P NP*
PP
NP
of
John
NP