Outline Grammar Formalisms Syntax in LTAG The derivation tree 1 - - PowerPoint PPT Presentation

Grammar Formalisms Syntax in LTAG

Laura Kallmeyer, Timm Lichte, Wolfgang Maier Universit¨ at T¨ ubingen 30.04.2007

Syntax in LTAG 1

Outline

1

The derivation tree

2

Design principles for elementary trees

3

Sample derivations

Syntax in LTAG 2

Derivation trees (1)

TAG derivations are described by derivation trees: For each derivation in a TAG there is a corresponding derivation

• tree. This tree contains

nodes for all elementary trees used in the derivation, and edges for all adjunctions and substitutions performed throughout the derivation. Whenever an elementary tree γ was attached to the node at address p in the elementary tree γ′, there is an edge from γ′ to γ labeled with p. (For the tree addresses, Gorn addresses are used: The root has address ǫ (or 0), and the ith daughter of the node with address p has address pi.)

Syntax in LTAG 3

Derivation trees (2)

Example: derivation tree for the derivation of John sometimes laughs laughs

1 2

john sometimes ⇒ Semantic dependency graph? Derivation trees are context-free, and uniquely determine the derived tree. ⇒ TAG is a linear context-free rewriting system, LCFRS

Syntax in LTAG 4

Linguistic analyses with LTAG

Linguistic role of the adjunction/substitution dualism Design principles for elementary trees

Lexicalization Extended domain of locality Factoring of recursion Elementary tree minimality

Syntax in LTAG 5

Lexicalization

Each elementary tree has at least one non-empty lexical item, its lexical anchor. ⇒ All widely used grammar formalisms support some kind of lexicalization! Reasons for lexicalization: The properties of a constituent depend on the lexical items

• ccurring in the constituent: the structure of a VP depends
• n the subcategorization properties of its verb.

VP → V intrans VP → V transNP

S NPs VP V sleeps S NPs VP V NPo likes

Syntax in LTAG 6

Elementary tree minimality

Elementary trees contain slots only for the complements of their lexical head. Elementary trees are projections of lexical items. Factoring of recursion Example: to-infinitives Example: (1) John gives a book to Mary S NP↓ VP V NP↓ PP gives P NP↓ to

Syntax in LTAG 7

Extended domain of locality

All dependencies are represented in elementary trees. locality constraints (constraints for UDC, island constraints) → adjunction constraints (2)

• a. whoi did John tell Sam that Bill likes ti
• b. whoi did John tell Sam that Mary said that Bill likes ti

S WHi SOA COMP S that NP VP WHi V NP who NP likes ǫi Bill S INFL NP VP did V NP S∗ NP tell John NP Sam

Syntax in LTAG 8

Principles related to semantics

Predicate argument cooccurrence: Each elementary tree associated with a predicate contains substitution nodes for each of its arguments. Semantic anchoring: Elementary trees are not semantically void (to, that.) Compositional principle: An elementary tree corresponds to a single semantic unit.

Syntax in LTAG 9

Functional elements

Besides lexical predicates, there are functional elements (complementizers, determiners, auxiliaries, negation) whose treatment in LTAG is less clear. They can be either in separate elementary trees (e.g., XTAG grammar)

• r in the elementary tree of the lexical item they are

associated with.

Syntax in LTAG 10

Sample derivations

Complements: NPs, PPs, adjectives, clauses (raising, controlling) Adjuncts: adjectives, particles, relative clauses Complements ⇒ Substitution, Adjuncts ⇒ Adjunction ?

Syntax in LTAG 11

Sample derivations - Complements (1)

(3) John buys Bill a book Elementary trees: NP John S NP↓ VP V NP↓ NP↓ buys NP Bill NP Det N a book Derivation tree: buys

1 22 23

John Bill a book

Syntax in LTAG 12

Sample derivations - Complements (2)

(4) Bill hopes that John wins NP Bill S NP↓ VP V S∗ hopes S Comp S that NP↓ VP V wins NP John Derivation tree: wins

ǫ 1

hopes John

1

Bill

Syntax in LTAG 13

Sample derivations - Complements (2)

(5) John seems to like Bill VP V VP∗ seems S NP↓ VP VP NP↓ to like Derivation tree: to like

1 2 22

John seems Bill

Syntax in LTAG 14

Sample derivations - Complements (2)

(6) John expects [ Bill to win ] S NP↓ VP V S∗ expects S NP↓ VP to win Derivation tree: to win

ǫ 1

expects Bill

1

John

Syntax in LTAG 15

Sample derivations - Complements (2)

(7) John expected [Mary to make a comment] expected selects for a subject NP and an infinitival sentence: NP John S NP↓ VP V S∗ expected S NP↓ VP to make a comment The sentential object is realised as a foot node in order to allow extractions: (8) whom does John expect to come?

Syntax in LTAG 16

Sample derivations - Complements (3)

to make a comment: make and comment in the same elementary tree since they form a light verb construction: S NP↓ VP V NP to make N comment NP Det NP∗ a

Syntax in LTAG 17

Sample derivations - Adjuncts (1)

(9) the good student participated in every course during the semester N AP N∗ A good NP Det N the student S NP↓ VP V PP participated P NP↓ in VP VP∗ PP P NP↓ during

Syntax in LTAG 18

Sample derivations - Adjuncts (1)

(10) the dog [who ate the cake] Extraposed relative clauses: (11) Somebodyi lives nearby [whoi has a CD-burner].

Syntax in LTAG 19

Derivation trees = Semantic dependency structure ?

The derivation tree is not always the semantic dependency structure: (12) John claims Bill is likely to win to win

1 ǫ 2

Bill claims is likely

1

John

Syntax in LTAG 20

Symmary

TAG derivations are described by derivation trees. In LTAG, elementary trees for lexical predicates contain slots for all arguments of these predicates, for nothing else. Recursion is factored away. The derived tree describes the constituent structure while the derivation tree is close to a semantic dependency graph.

Syntax in LTAG 21