Semantic Modeling with Frames Rainer Osswald & Wiebke Petersen - - PowerPoint PPT Presentation

semantic modeling with frames
SMART_READER_LITE
LIVE PREVIEW

Semantic Modeling with Frames Rainer Osswald & Wiebke Petersen - - PowerPoint PPT Presentation

Semantic Modeling with Frames Rainer Osswald & Wiebke Petersen Department of Linguistics and Information Science Heinrich-Heine-Universit at D usseldorf ESSLLI 2018 Introductory Course Sofia University 06. 08. 10. 08. 2018 SFB


slide-1
SLIDE 1

Semantic Modeling with Frames

Rainer Osswald & Wiebke Petersen

Department of Linguistics and Information Science Heinrich-Heine-Universit¨ at D¨ usseldorf

ESSLLI 2018

Introductory Course

Sofia University

  • 06. 08. – 10. 08. 2018

SFB 991

slide-2
SLIDE 2

Part 3 A model of the syntax-semantics interface

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 1

slide-3
SLIDE 3

Topics

Overview of the approach Lexicalized Tree Adjoining Grammars (LTAG) Feature structures based TAG (FTAG) Tree families and factorization in the metagrammar Elementary construction = elementary tree + semantic frame Applications (directed motion constructions, ...) Outlook: Factorization of constructions

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 2

slide-4
SLIDE 4

Introduction

Example

(1) Adam ate an apple.

S VP[I=e] NP[I=y] V ‘ate’ NP[I=x] e         eating actor x theme y        

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 3

slide-5
SLIDE 5

Introduction

Example

(1) Adam ate an apple.

NP[I=u] ‘Adam’ u person name ‘Adam’

  • S

VP[I=e] NP[I=y] V ‘ate’ NP[I=x] e         eating actor x theme y         NP[I=v] ‘an apple’ v

  • apple
  • x u

y v

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 3

slide-6
SLIDE 6

Introduction

Example

(1) Adam ate an apple.

NP[I=u] ‘Adam’ u person name ‘Adam’

  • S

VP[I=e] NP[I=y] V ‘ate’ NP[I=x] e         eating actor x theme y         NP[I=v] ‘an apple’ v

  • apple
  • x u

y v

S VP[I=e] NP[I=y] ‘an apple’ V ‘ate’ NP[I=x] ‘Adam’ e             eating actor x       person name ‘Adam’       theme y

  • apple

           e eating x person ‘Adam’ y apple actor name theme

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 3

slide-7
SLIDE 7

Introduction

Overview of the approach

Semantic composition (≈ unification) is triggered by syntactic composition (≈ substitution and adjunction). Semantic representations are linked to entire elementary trees. (A further decomposition is possible in the “metagrammar”.) Interface features relate nodes in the syntactic tree to components in the semantic representation.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 4

slide-8
SLIDE 8

Introduction

Overview of the approach

Semantic composition (≈ unification) is triggered by syntactic composition (≈ substitution and adjunction). Semantic representations are linked to entire elementary trees. (A further decomposition is possible in the “metagrammar”.) Interface features relate nodes in the syntactic tree to components in the semantic representation.

Main components of the framework

[Kallmeyer/Osswald 2013]

Frame Semantics Lexicalized Tree Adjoining Grammars (LTAG)

[Joshi/Schabes 1997; Abeille/Rambow 2000]

Metagrammatical specification and decomposition

[Crabb´ e/Duchier 2005; Crabb´ e et al. 2013, Lichte/Petitjean 2015]

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 4

slide-9
SLIDE 9

Lexicalized Tree Adjoining Grammars (LTAG)

Tree-rewriting system; mildly context sensitive grammar formalism Finite set of (lexicalized) elementary trees. Two operations: substitution (replacing a leaf with a new tree) and adjunction (replacing an internal node with a new tree).

NP ‘Adam’ S VP NP V ‘ate’ NP NP ‘an apple’ VP VP∗ Adv ‘always’

  • S

VP VP NP ‘an apple’ V ‘ate’ Adv ‘always’ NP ‘Adam’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 5

slide-10
SLIDE 10

Lexicalized Tree Adjoining Grammars (LTAG)

Feature-structure based TAG (FTAG)

[Vijay-Shanker/Joshi 1988]

Each node has a top and a botom feature structure: The top feature structure provide information about what the node presents within the surrounding structure. The botom feature structure provide information about what the tree below the node represents. In the final derived tree, top and botom must be unified. Operations on feature structures under substitution: The top of the root of the new initial tree unifies with the top of the substitution node. Operations on feature structures under adjunction: The top of the root of the new auxiliary tree unifies with the top of the adjunction site; the botom of the foot of the new tree unifies with the botom of the adjunction site.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 6

slide-11
SLIDE 11

Lexicalized Tree Adjoining Grammars (LTAG)

Example

NP[]

[agr=[pers=3, num=sg]]

‘John’ S NP[agr= 1 ] VP[agr= 1 , mode=ind]

[mode=ger]

V ‘singing’ VP[agr= 2 , mode= 3 ] V[mode= 3 ind]

[agr= 2 [pers=3, num=sg]]

VP∗[mode=ger] ‘is’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 7

slide-12
SLIDE 12

Lexicalized Tree Adjoining Grammars (LTAG)

Example

NP[]

[agr=[pers=3, num=sg]]

‘John’ S NP[agr= 1 ] VP[agr= 1 , mode=ind]

[mode=ger]

V ‘singing’ VP[agr= 2 , mode= 3 ] V[mode= 3 ind]

[agr= 2 [pers=3, num=sg]]

VP∗[mode=ger] ‘is’

Result of derivation:

S NP[agr= 1 ]

[agr=[pers=3, num=sg]]

VP[agr= 1 ,mode=ind]

[agr= 2 , mode= 3 ]

‘John’ V[mode= 3 ind]

[agr= 2 [pers=3, num=sg]]

VP[mode=ger]

[mode=ger]

‘is’ V ‘singing’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 7

slide-13
SLIDE 13

Lexicalized Tree Adjoining Grammars (LTAG)

Example

NP[]

[agr=[pers=3, num=sg]]

‘John’ S NP[agr= 1 ] VP[agr= 1 , mode=ind]

[mode=ger]

V ‘singing’ VP[agr= 2 , mode= 3 ] V[mode= 3 ind]

[agr= 2 [pers=3, num=sg]]

VP∗[mode=ger] ‘is’

Afer top-botom unifications:

S NP[agr = 1 ] VP[agr = 1 [pers = 3, num =sg]

mode =ind ]

‘John’ V[agr = 1

mode=ind]

VP[mode =ger] ‘is’ V ‘singing’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 7

slide-14
SLIDE 14

Lexicalized Tree Adjoining Grammars (LTAG)

Two key properties of the LTAG formalism Extended domain of locality The full argument projection of a lexical item can be represented by a single elementary tree. Elementary trees can have a complex constituent structure. Factoring recursion from the domain of dependencies Constructions related to iteration and recursion are modeled by adjunction. Through adjunction, the local dependencies encoded by elementary trees can become long-distance dependencies in the derived trees.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 8

slide-15
SLIDE 15

Lexicalized Tree Adjoining Grammars (LTAG)

Two key properties of the LTAG formalism Extended domain of locality The full argument projection of a lexical item can be represented by a single elementary tree. Elementary trees can have a complex constituent structure. Factoring recursion from the domain of dependencies Constructions related to iteration and recursion are modeled by adjunction. Through adjunction, the local dependencies encoded by elementary trees can become long-distance dependencies in the derived trees. Slogan: “Complicate locally, simplify globally”

[Bangalore/Joshi 2010]

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 8

slide-16
SLIDE 16

Lexicalized Tree Adjoining Grammars (LTAG)

“Simplify globally” The composition of elementary trees can be expressed by two general operations: substitution and adjunction. (Since basically all linguistic constraints are specified over the local domains represented by elementary trees.) “Complicate locally” Elementary trees can have complex semantic representations which are not necessarily derived compositionally (in the syntax) from smaller parts of the trees. In particular, there is no need to reproduce the internal structure

  • f an elementary syntactic tree within its associated semantic

representation.

[Kallmeyer/Joshi 2003]

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 9

slide-17
SLIDE 17

Lexicalized Tree Adjoining Grammars (LTAG)

Tree families Unanchored elementary trees are organized in tree families, which capture variations in the (syntactic) subcategorization frames. Example unanchored family for transitive verbs

S NP VP V◇ NP S NP S NP VP ε V◇ NP S NP VP V◇ PP P NP by S NP S NP VP ε V◇ PP P NP by S NP S NP VP V◇ NP ε ...

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 10

slide-18
SLIDE 18

Lexicalized Tree Adjoining Grammars (LTAG)

Tree families Unanchored elementary trees are organized in tree families, which capture variations in the (syntactic) subcategorization frames. Example unanchored family for transitive verbs

S NP VP V◇ NP S NP S NP VP ε V◇ NP S NP VP V◇ PP P NP by S NP S NP VP ε V◇ PP P NP by S NP S NP VP V◇ NP ε ...

Metagrammar Modular characterization of elementary trees by a system of tree descriptions.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 10

slide-19
SLIDE 19

LTAG & metagrammar specification

Decomposition/factorization in the metagrammar

Class CanSubj S NP ≺ VP V◇ Class ExtrSubj S NP[wh=yes] ≺∗ S NP ≺ VP ε V◇ Class Subj CanSubj ∨ ExtSubj Class DirObj VP V◇ ≺∗ NP Class ByObj VP[voice=passive] V◇ ≺∗ PP P ≺ NP by Class ActV VP[voice=active] V◇ Class PassV VP[voice=passive] V◇ Class Transitive ((Subj ∧ ActV ) ∨ ByObj ∨ PassV ) ∧ ((DirObj ∧ ActV ) ∨ (Subj ∧ PassV ))

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 11

slide-20
SLIDE 20

LTAG & metagrammar specification

XMG (eXtensible MetaGrammar)

[e.g., Crabb´ e et al. 2013]

A framework for specifying (the elementary structures of) tree based grammars by means of a declarative language (e.g., by dominance and precedence constraints) The specifications are organized into classes that can be reused (“imported”) by other classes. Classes may contain descriptions from different dimensions, and the XMG system can be extended in this respect, e.g., by a dimension of frame descriptions. An XMG compiler generates the elementary structures of a grammar from a metagrammar.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 12

slide-21
SLIDE 21

LTAG & metagrammar specification

Example

Class CanSubj S NP ≺ VP V◇ Class ExtrSubj S NP[wh=yes] ≺∗ S NP ≺ VP ε V◇ Class Subj CanSubj ∨ ExtSubj Class DirObj VP V◇ ≺∗ NP Class ByObj VP[voice=passive] V◇ ≺∗ PP P ≺ NP by Class ActV VP[voice=active] V◇ Class PassV VP[voice=passive] V◇ Class Transitive ((Subj ∧ ActV ) ∨ ByObj ∨ PassV ) ∧ ((DirObj ∧ ActV ) ∨ (Subj ∧ PassV ))

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 13

slide-22
SLIDE 22

LTAG & metagrammar specification

Summary of the LTAG architecture metagrammar classes compilation unanchored tree families lexical entries lexical selection LTAG

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 14

slide-23
SLIDE 23

LTAG & metagrammar specification

Summary of the LTAG architecture metagrammar classes compilation unanchored tree families lexical entries lexical selection LTAG Next step: Add (frame) semantics to all components and link syntax to semantics.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 14

slide-24
SLIDE 24

LTAG & frame semantics

Overall architecture (syntax + semantics) metagrammar classes + AV constraints compilation unanchored families

  • f constructions

lexical entries (+ frame semantics) lexical selection LTAG + frames

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 15

slide-25
SLIDE 25

LTAG & frame semantics

Elements of the syntax-semantics interface Elementary construction: elementary tree + semantic frame + linking of frame node variables to interface features in the tree Specification in the metagrammar: classes of tree constraints + sets of atribute-value constraints + linking of variables to interface features Note: Regularities about argument linking are expressed in the metagrammar.

[Kallmeyer/Lichte/Osswald/Petitjean 2016]

Semantic composition ≈ frame unification via identification of interface variables during substitution and adjunction.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 16

slide-26
SLIDE 26

Applications: directed motion construction

Intransitive: (2) a. Mary walked to the house.

  • b. The ball rolled into the goal.
  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 17

slide-27
SLIDE 27

Applications: directed motion construction

Intransitive: (2) a. Mary walked to the house.

  • b. The ball rolled into the goal.

Transitive: (3) a. John threw/kicked the ball into the goal.

  • b. John pushed/pulled the cart to the station.
  • c. John rolled the ball into the hole.
  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 17

slide-28
SLIDE 28

Applications: directed motion construction

Intransitive: (2) a. Mary walked to the house.

  • b. The ball rolled into the goal.

Transitive: (3) a. John threw/kicked the ball into the goal.

  • b. John pushed/pulled the cart to the station.
  • c. John rolled the ball into the hole.

Directional specifications are not restricted to goal expressions but can also describe the source or the course of the path in more detail.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 17

slide-29
SLIDE 29

Applications: directed motion construction

Intransitive: (2) a. Mary walked to the house.

  • b. The ball rolled into the goal.

Transitive: (3) a. John threw/kicked the ball into the goal.

  • b. John pushed/pulled the cart to the station.
  • c. John rolled the ball into the hole.

Directional specifications are not restricted to goal expressions but can also describe the source or the course of the path in more detail. Moreover, path descriptions can be iterated to some extent: (4) a. John walked through the gate along the fence to the house.

  • b. John threw the ball over the fence into the yard.
  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 17

slide-30
SLIDE 30

Applications: directed motion construction

Qestion: Syntactic treatment of directional PPs ? Construction ( elementary tree) Syntactic composition ( adjunction)

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 18

slide-31
SLIDE 31

Applications: directed motion construction

Qestion: Syntactic treatment of directional PPs ? Construction ( elementary tree) Syntactic composition ( adjunction) Arguments for treating goal (or bounded) PPs constructionally, in contrast to path (or unbounded) PPs: Goal PPs cannot be iterated.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 18

slide-32
SLIDE 32

Applications: directed motion construction

Qestion: Syntactic treatment of directional PPs ? Construction ( elementary tree) Syntactic composition ( adjunction) Arguments for treating goal (or bounded) PPs constructionally, in contrast to path (or unbounded) PPs: Goal PPs cannot be iterated. They affect the Aktionsart of the expression: (5) a. She walked (*in half an hour/for half an hour).

  • b. She walked to the brook (in half an hour/*for half an hour).
  • c. She walked along the brook (*in half an hour/for half an hour).
  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 18

slide-33
SLIDE 33

Applications: directed motion construction

Unanchored construction for intransitive directed motion (n0Vpp(dir)): S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V◇[e=e]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation mover x goal z path [path] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 19

slide-34
SLIDE 34

Applications: directed motion construction

Unanchored construction for intransitive directed motion (n0Vpp(dir)): S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V◇[e=e]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation mover x goal z path [path] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

Elementary tree for ‘into’: PP[i=z,e=e] P NP[i=z]

‘into’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation path [path endp

1 ]

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ z [in-region

2 ]

part-of( 1 , 2 )

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 19

slide-35
SLIDE 35

Applications: directed motion construction

Example (intransitive directed motion)

(6) John walked into the house.

NP[i=x′] ‘John’

x′[person name ‘John’]

S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V[e=e] ‘walked’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-locomotion actor

1 x

mover

1

goal z path [path] manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ event path [path endp

1 ]

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ z′[in-region

2 ]

part-of( 1 , 2 )

PP[i=z′,e=e′] P NP[i=z′] ‘into’ NP[i=z′′] Det N ‘the’ ‘house’

z′′[house in-region [region]]

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 20

slide-36
SLIDE 36

Applications: directed motion construction

Example (intransitive directed motion)

(6) John walked into the house.

NP[i=x′] ‘John’

x′[person name ‘John’]

S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V[e=e] P NP[i=z] ‘walked’ ‘into’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-locomotion actor

1 x

mover

1

goal z [in-region

3 ]

path [path endp

2 ]

manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ part-of( 2 , 3 ) z′′[house in-region [region]]

NP[i=z′′] Det N ‘the’ ‘house’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 20

slide-37
SLIDE 37

Applications: directed motion construction

Example (intransitive directed motion)

(6) John walked into the house.

S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V[e=e] P NP[i=z] Det N ‘John’ ‘walked’ ‘into’ ‘the’ ‘house’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-locomotion actor

1 x [person

name ‘John’] mover

1

goal z [house in-region

3 [region]]

path [path endp

2 ]

manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ part-of( 2 , 3 )

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 20

slide-38
SLIDE 38

Applications: directed motion construction

Example (intransitive directed motion)

(6) John walked into the house.

S NP[i=x] VP[e=e] VP[e=e] PP[i=z,e=e] V[e=e] P NP[i=z] Det N ‘John’ ‘walked’ ‘into’ ‘the’ ‘house’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-locomotion actor

1 x [person

name ‘John’] mover

1

goal z [house in-region

3 [region]]

path [path endp

2 ]

manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ part-of( 2 , 3 ) e bounded-locomotion x person path walking ‘John’ z house region actor mover path manner name endp goal in-region part-of

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 20

slide-39
SLIDE 39

Applications: directed motion construction

Lexical anchoring (non-directed case)

morph entry ‘walked’ pos: V Syn1:

⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ agr = ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ pers = 3 num = sg ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

lemma: walk + lemma entry walk: FAM: n0V, ... Syn2:

[e = e0]

Sem :

e0 ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ locomotion manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎦

+ Constraints: locomotion ⇛ activity ∧ translocation translocation ⇛ motion ∧ path ∶ path activity ⇛ actor ∶ ⊺ motion ⇛ mover ∶ ⊺ activity ∧ motion ⇛ actor ≐ mover

e0 ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ locomotion actor

1

mover

1

path [path] manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

V[agr = ...,e = e0] ‘walked’ S NP[i=x] VP[e=e] V◇[e=e]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ activity actor x ⎤ ⎥ ⎥ ⎥ ⎥ ⎦ e0 ≜e

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ locomotion actor

1 x

mover

1

path [path] manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

S NP[i=x] VP[e=e] V[agr = ...,e = e] ‘walked’

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 21

slide-40
SLIDE 40

Applications: directed motion construction

Example

(7) John walked along the brook.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 22

slide-41
SLIDE 41

Applications: directed motion construction

Example

(7) John walked along the brook. S NP[i=x] VP[e=e] V[e=e] walked

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ locomotion actor

1 x

mover

1

path [path] manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

VP VP∗[e=e′] PP[i=z] P NP[i=z] along

e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ translocation path [path region

2 ]

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ z [at-region

3 ]

part-of( 2 , 3 )

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 22

slide-42
SLIDE 42

Applications: directed motion construction

Example

(7) John walked along the brook. S NP[i=x] VP[e=e] V[e=e] walked

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ locomotion actor

1 x

mover

1

path [path] manner [walking] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

VP VP∗[e=e′] PP[i=z] P NP[i=z] along

e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ translocation path [path region

2 ]

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ z [at-region

3 ]

part-of( 2 , 3 ) e locomotion x person path walking region z region actor mover path manner region at-region part-of

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 22

slide-43
SLIDE 43

Applications: directed motion construction

Example (causative directed motion)

(8) Mary threw/kicked/rolled the ball into the room.

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 23

slide-44
SLIDE 44

Applications: directed motion construction

Example (causative directed motion)

(8) Mary threw/kicked/rolled the ball into the room. Unanchored construction (n0Vn1pp(dir)):

S NP[i=x] VP[e=e] VP[e = e,path = p] PP[i = z,e = e′] V◇[e=e] NP[i=y]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ causation cause ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ activity actor x theme y ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ effect e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation mover y goal z path p ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 23

slide-45
SLIDE 45

Applications: directed motion construction

Example (causative directed motion)

(8) Mary threw/kicked/rolled the ball into the room. Unanchored construction (n0Vn1pp(dir)):

S NP[i=x] VP[e=e] VP[e = e,path = p] PP[i = z,e = e′] V◇[e=e] NP[i=y]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ causation cause ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ activity actor x theme y ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ effect e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation mover y goal z path p ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

(Partial) lexical entry for ‘threw’: V[e=e]

‘threw’

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣

  • nset-causation

cause ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ activity actor ⊺ theme

1

manner [throwing] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ effect ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ translocation ∧ undergoing theme

1

mover

1

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 23

slide-46
SLIDE 46

Applications: directed motion construction

Outlook: Metagrammar classes (syntax + semantics)

✎ ✍ ☞ ✌

Class n0Vn1

✎ ✍ ☞ ✌

Class n0V

✎ ✍ ☞ ✌

Class Subj S NP[agr = 1 ,i = x] ≺ VP[agr= 1 ] V◇[e = e]

e[event actor x] ✎ ✍ ☞ ✌

Class VSpine VP[agr = 1 ,e = 2 ] V◇[agr = 1 ,e = 2 ]

✎ ✍ ☞ ✌

Class DirObj VP V◇[e = e] ≺∗ NP[i = x]

e[event theme x] ∨ ...

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 24

slide-47
SLIDE 47

Applications: directed motion construction

Outlook: Metagrammar classes (syntax + semantics)

✎ ✍ ☞ ✌ Class DirPrepObj export: e, x VP[path=p] VP[path=p] ≺ PP[i=x,e=e] V◇

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation goal x path p ⎤ ⎥ ⎥ ⎥ ⎥ ⎦

✎ ✍ ☞ ✌ Class n0Vpp(dir) identities: C1.e = C2.e ✎ ✍ ☞ ✌ Class C1 =n0V export: e ... ✎ ✍ ☞ ✌ Class C2 =DirPrepObj export: e, x ... ✎ ✍ ☞ ✌ Class n0Vn1pp(dir) identities: C1.e = e, C2.x = z, C2.e = e′ ✎ ✍ ☞ ✌ Class C1 =n0Vn1 export: e ... ✎ ✍ ☞ ✌ Class C2 =DirPrepObj export: e, x ...

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ causation actor x 1 theme y 2 cause ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ activity actor

1

theme

2

⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ effect e′[mover

2

goal z ] ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 25

slide-48
SLIDE 48

Applications: dative alternation

Sketch

[→ Kallmeyer/Osswald 2013]

(9) a. John sent Mary the book. (double object construction)

  • b. John sent the book to Mary.

(prepositional object construction)

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 26

slide-49
SLIDE 49

Applications: dative alternation

Sketch

[→ Kallmeyer/Osswald 2013]

(9) a. John sent Mary the book. (double object construction)

  • b. John sent the book to Mary.

(prepositional object construction) a) S NP[i=x] VP[e=e] V◇[e=e] NP[i=z] NP[i=y]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ causation cause [activity actor x] effect ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ change-of-possession theme y recipient z ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

b) S NP[i=x] VP[e=e] VP[e=e] PP[prep = to,i = z,e = e′] V◇[e=e] NP[i=y]

e ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ causation cause [activity actor x] effect e′ ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ bounded-translocation mover y goal z ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 26

slide-50
SLIDE 50

XMG implementation of frame semantics

Examples

[from Lichte/Petitjean 2015]

<syn>-dimension of class Subj

  • [= 0 ]

⋄[= 0 ] [= 1 ]

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 27

slide-51
SLIDE 51

XMG implementation of frame semantics

Examples

[from Lichte/Petitjean 2015]

<syn>-dimension + <frame>-dimension of class Subj

  • [= 0 ]

⋄[= 0 ] [= 1 ]  

  • 1

 

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 28

slide-52
SLIDE 52

XMG implementation of frame semantics

Examples

[from Lichte/Petitjean 2015]

Specification of frames:

                 

  • 1
  • 2

 

  • 1
  • 2

  

  • 4

  2

  • 3

                   

  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 29

slide-53
SLIDE 53

XMG implementation of frame semantics

Examples

[from Lichte/Petitjean 2015]

Specification of frames:

                 

  • 1
  • 2

 

  • 1
  • 2

  

  • 4

  2

  • 3

                   

  • Specification of atribute-value constraints:
  • R. Osswald & W. Petersen

Semantic Modeling with Frames | Part 3 | ESSLLI 2018 | Sofia 29