A Best-Fit Approach for Productive Analysis of Omitted Arguments

1
A Best-Fit Approach for Productive Analysis of
Omitted Arguments

• Eva Mok John Bryant
• University of California, Berkeley
• International Computer Science Institute

2
Simplify grammar by exploiting the language
understanding process

• Omission of arguments in Mandarin Chinese
• Construction grammar framework
• Model of language understanding
• Our best-fit approach

3
Productive Argument Omission (in Mandarin)

• Mother (I) give you this (a toy).

ma1ma gei3 ni3 zhei4ge
mother give 2PS thisCLS
1

• You give auntie the peach.

2
ni3 gei3 yi2
2PS give auntie

• Oh (go on)! You give auntie that.

3
ao ni3 gei3 ya
EMP 2PS give EMP
4
gei3
give

• I give you some peach.

CHILDES Beijing Corpus (Tardiff, 1993 Tardiff,
1996)
4
Arguments are omitted with different probabilities

• All arguments omitted 30.6 No arguments
  omitted 6.1

5
Construction grammar approach

• Kay Fillmore 1999 Goldberg 1995
• Grammaticality form and function
• Basic unit of analysis construction, i.e. a
  pairing of form and meaning constraints
• Not purely lexically compositional
• Implies early use of semantics in processing
• Embodied Construction Grammar (ECG) (Bergen
  Chang, 2005)

6
Problem Proliferation of constructions
Subj Verb Obj1 Obj2
? ? ? ?
Giver Transfer Recipient Theme
Verb Obj1 Obj2
? ? ?
Transfer Recipient Theme
Subj Verb Obj2
? ? ?
Giver Transfer Theme
Subj Verb Obj1
? ? ?
Giver Transfer Recipient

7
If the analysis process is smart, then...
Subj Verb Obj1 Obj2
? ? ? ?
Giver Transfer Recipient Theme

• The grammar needs only state one construction
• Omission of constituents is flexibly allowed
• The analysis process figures out what was omitted

8
Best-fit analysis process takes burden off the
grammar representation
Constructions
Utterance
incremental, competition-based,
psycholinguistically plausible
Semantic Specification image schemas, frames,
action schemas
Simulation
9
Competition-based analyzer finds the best analysis

• An analysis is made up of
• A constructional tree
• A set of resolutions
• A semantic specification

The best fit has the highest combined score
10
Combined score that determines best-fit

• Syntactic Fit
• Constituency relations
• Combine with preferences on non-local elements
• Conditioned on syntactic context
• Antecedent Fit
• Ability to find referents in the context
• Conditioned on syntactic information, feature
  agreement
• Semantic Fit
• Semantic bindings for frame roles
• Frame roles fillers are scored

11
Analyzing ni3 gei3 yi2 (You give auntie)
Two of the competing analyses
ni3 gei3 yi2 omitted
? ? ? ?
Giver Transfer Recipient Theme
ni3 gei3 omitted yi2
? ? ? ?
Giver Transfer Recipient Theme

• Syntactic Fit
• P(Theme omitted ditransitive cxn) 0.65
• P(Recipient omitted ditransitive cxn) 0.42

(1-0.78)(1-0.42)0.65 0.08
(1-0.78)(1-0.65)0.42 0.03
12
Using frame and lexical information to restrict
type of reference
The Transfer Frame Giver Recipient Theme The Transfer Frame Giver Recipient Theme
Manner Means Place Purpose Reason Time
Lexical Unit gei3 Giver (DNI) Recipient (DNI) Theme (DNI)
13
Can the omitted argument be recovered from
context?

• Antecedent Fit

ni3 gei3 yi2 omitted
? ? ? ?
Giver Transfer Recipient Theme
ni3 gei3 omitted yi2
? ? ? ?
Giver Transfer Recipient Theme
?
14
How good of a theme is a peach? How about an
aunt?

• Semantic Fit

ni3 gei3 yi2 omitted
? ? ? ?
Giver Transfer Recipient Theme
ni3 gei3 omitted yi2
? ? ? ?
Giver Transfer Recipient Theme
The Transfer Frame Giver (usually animate) Recipient (usually animate) Theme (usually inanimate)
15
The argument omission patterns shown earlier can
be covered with just ONE construction
Subj Verb Obj1 Obj2
? ? ? ?
Giver Transfer Recipient Theme
P(omittedcxn)
0.78
0.42
0.65

• Each cxn is annotated with probabilities of
  omission
• Language-specific default probability can be set

16
Research goal

• A computationally-precise modeling framework for
  learning early constructions

Learning
New Construction
Linguistic Knowledge
17
Frequent argument omission in pro-drop languages

• Mandarin example
• ni3 gei3 yi2 (you give auntie)
• Even in English, there are often no spoken
  antecedents to pronouns in conversations

Learner must integrate cues from intentions,
gestures, prior discourse, etc
18
A short dialogue

• bie2 mo3 wai4tou2 a 1_3 ! (?????)
• NEG-IMP apply forehead
• Dont apply lotion to your forehead
• mo3 wai4tou2 ke3 jiu4 bu4 hao3kan4 le a .
  (??????????)
• apply forehead LINKER LINKER NEG good looking CRS
  SFP
• If you apply lotion to your forehead then
  you will not be pretty
• ze ya a bie2 gei3 ma1ma wang3 lian3 shang4
  mo3 e ! (??? ?????????)
• INTERJ NEG-IMP BEN mother CV-DIR face on apply
• INTERJ Dont apply the lotion on your moms
  face (for mom)
• - low pitch motherese ma1ma bu4 mo3 you2 .
  (?????)
• mother NEG apply lotion
• Mom doesnt apply (use) lotion

19
Goals, refined

• Demonstrate learning given
• embodied meaning representation
• structured representation of context
• Based on
• Usage-based learning
• Domain-general statistical learning mechanism
• Generalization / linguistic category formation

20
Towards a precise computational model

• Modeling early grammar learning
• Context model Simulation
• Data annotation
• Finding the best analysis for learning
• Hypothesizing and reorganizing constructions
• Pilot results

21
Embodied Construction Grammar
construction yi2-N subcase of Morpheme
form constraints self.f.orth lt--
"yi2" meaning _at_Aunt evokes RD as
rd constraints self.m lt--gt
rd.referent self.m lt--gt rd.ontological_category
22
you specifies discourse role
construction ni3-N subcase of Morpheme
form constraints self.f.orth lt--
"ni3" meaning _at_Human evokes RD as
rd constraints self.m lt--gt
rd.referent self.m lt--gt rd.ontological_category
rd.discourse_participant_role lt--
_at_Addressee rd.set_size lt-- _at_Singleton
23
The meaning of give is a schema with roles
schema Transfer subcase of Action roles giver
_at_Entity recipient _at_Entity theme
_at_Entity constraints giver lt--gt protagonist
construction gei3-V2 subcase of Morpheme
form constraints self.f.orth lt--
"gei3" meaning Give
schema Give subcase of Transfer constraints in
herent_aspect lt-- _at_Inherent_Achievement giver
lt-- _at_Animate recipient lt-- _at_Animate theme lt--
_at_Manipulable_Inanimate_Object
24
Finally, you-give-aunt links up the roles
construction ni3-gei3-yi2 subcase of
Finite_Clause constructional constituents n
ni3-N g gei3-V2 y yi2-N form const
raints n.f meets g.f g.f meets y.f meaning
Give constraints self.m lt--gt
g.m self.m.giver lt--gt n.m self.m.recipient
lt--gt y.m
25
The learning loop Hypothesize Reorganize
World Knowledge
Utterance
Linguistic Knowledge
reorganize
Analysis
reinforcement
Context Fitting
hypothesize
PartialSemSpec
26
If the learner has a ditransitive cxn
Context
Meaning
Form
XIXI
MOT
addressee
speaker
giver
meets
Discourse Segment
meets
recipient
INV
theme
Peach
27
Context fitting recovers more relations
Context
Meaning
Form
XIXI
MOT
addressee
speaker
giver
meets
giver
Discourse Segment
Give
recipient
meets
recipient
attentional-focus
INV
theme
theme
Peach
28
But the learner does not yet have phrasal cxns
Context
Meaning
Form
XIXI
Give
INV
29
Context bootstraps learning
Meaning
Form
construction ni3-gei3-yi2 subcase of
Finite_Clause constructional constituents n
ni3 g gei3 y yi2 form constraints
n.f meets g.f g.f meets y.f meaning
Give constraints self.m lt--gt
g.m self.m.giver lt--gt n.m self.m.recipient
lt--gt y.m
30
A model of context is key to learning

• The context model makes it possible for the
  learning model to
• learn new constructions using contextually
  available information
• learn argument-structure constructions in
  pro-drop languages

31
Understanding an utterance in context
Transcripts
Schemas Constructions
Analysis Resolution
Context Fitting
Simulation
Semantic Specification
32
Context model Events Utterances
Setting participants, entities, relations
Start
Event
Event
DS
Sub-Event
Sub-Event
33
Entities and Relations are instantiated
Setting CHI, MOT (incl. body parts) livingroom (
incl. ground, ceiling, chair, etc), lotion
ds04 admonishing05 speaker MOT addressee
CHI forcefulness normal
caused_motion01 forceful_motion motion
Start
apply02 applier CHI substance lotion surface
face(CHI)
translational_motion03 mover lotion spg SPG
34
The context model is updated dynamically

• Extended transcript annotation speech acts
  events
• Simulator inserts events into context model
  updates it with the effects
• Some relations persists over time some dont.

Simulation
35
Competition-based analyzer finds the best analysis

• An analysis is made up of
• A constructional tree
• A semantic specification
• A set of resolutions

Bill gave Mary the book
36
Combined score that determines best-fit

• Syntactic Fit
• Constituency relations
• Combine with preferences on non-local elements
• Conditioned on syntactic context
• Antecedent Fit
• Ability to find referents in the context
• Conditioned on syntactic information, feature
  agreement
• Semantic Fit
• Semantic bindings for frame roles
• Frame roles fillers are scored

37
Context Fitting goes beyond resolution
Context
Meaning
Form
XIXI
MOT
addressee
speaker
giver
meets
giver
Discourse Segment
Give
recipient
meets
recipient
attentional-focus
INV
theme
theme
Peach
38
Context Fitting, a.k.a. intention reading

• Context Fitting takes resolution a step further
• considers entire context model, ranked by recency
• considers relations amongst entities
• heuristically fits from top down, e.g.
• discourse-related entities
• complex processes
• simple processes
• other structured and unstructured entities
• more heuristics for future events (e.g. in cases
  of commands or suggestions)

39
Adult grammar size

• 615 constructions total
• 100 abstract cxns (26 to capture lexical
  variants)
• 70 phrasal/clausal cxns
• 440 lexical cxns (260 open class)
• 195 schemas (120 open class, 75 closed class)

40
Starter learner grammar size

• No grammatical categories (except interjections)
• Lexical items only
• 440 lexical constructions
• 260 open class schema / ontology meanings
• 40 closed class pronouns, negation markers, etc
• 60 function words no meanings
• 195 schemas (120 open class, 75 closed class)

41
The process hierarchy defined in schemas
Process
Proto_Transitive
State_Change
State
Action
Complex_Process
Intransitive_State
Two_Participant_State
Serial_Processes
Mental_State
Concurrent_Processes
Cause_Effect
Joint_Motion
Caused_Motion
42
The process hierarchy defined in schemas
Action
Motion
Intransitive_Action
Translational_Motion
Expression
Self_Motion
Translational_Self_Motion
Force_Application
Continuous_Force_Application
Forceful_Motion
Agentive_Impact
43
The process hierarchy defined in schemas
Action
Communication
Cause_Change
Obtainment
Transfer
Ingestion
Perception
Other_Transitive_Action
44
Understanding an utterance in context
Transcripts
reorganize
Schemas Constructions
Analysis Resolution
Context Fitting
Simulation
reinforcement
hypothesize
Semantic Specification
45
Hypothesize Reorganize

• Hypothesize
• utterance-driven
• relies on the analysis (SemSpec context)
• operations compose
• Reorganize
• grammar-driven
• can be triggered by usage (to be determined)
• operations generalize

46
Composing new constructions
Context
MOT
XIXI
Peach
giver recipient theme
INV

• Compose operation If roles from different
  constructions point to the same context element,
  propose a new construction and set up a meaning
  binding.

47
Creating pivot constructions

• Pivot generalization Given a phrasal cxn, look
  for another cxn that shares 1 constituents. Line
  up roles and bindings. Create new cxn category
  for the slot.

48
Resulting constructions
construction ni3-gei3-cat01 constituents ni3,
gei3, cat01 meaning Give constraints self.
m.recipient lt--gt g.m
general construction cat01 subcase of
Morpheme meaning _at_Human
construction wo3 subcase of cat01 meaning
_at_Human
construction yi2 subcase of cat01 meaning _at_Aunt
49
Pilot Results Sample constructions learned

• Composed
• Pivot Cxns

chi1_fan4 ni3_chuan1_xie2 ni3_shuo1 bu4_na2 wo3_qu4 ni3_ping2zi_gei3_wo3 ni3_gei3_yi2 wo3_bu4_chi1 eat rice you wear shoe you say NEG take I go you bottle give me you give auntI NEG eat
ni3 shuo1, chuan1 ni3 shuo1, hua4 wo3 zhao3, qu4 bu4 na2, he1wo3, ma1 cheng2 you say, wear you say, draw I find, go NEG take, drink I, mom scoop
50
Challenge 1 Non-compositional meaning
Context
Bake-Event
baker baked
MOT
CHI
Cake

• Non-compositional meaning Search for additional
  meaning schemas (in context or in general) that
  relate the meanings of the individual
  constructions

51
Challenge 2 Function words
Context
Bake-Event
baker baked
MOT
Cake
CHI

• Function words tend to indicate relations rather
  than events or entities

52
Challenge 3 How far up to generalize

• Eat rice
• Eat apple
• Eat watermelon
• Want rice
• Want apple
• Want chair

Inanimate Object
Manipulable Objects
Unmovable Objects
Food
Furniture
Fruit
Savory
Chair
Sofa
apple
watermelon
rice
53
Challenge 4 Beyond pivot constructions

• Pivot constructions indexing on particular
  constituent type
• Eat rice Eat apple Eat watermelon
• Abstract constructions indexing on role-filler
  relations between constituents

Schema Eat roles eater lt--gt agent food lt--gt
patient
Schema Want roles wanter lt--gt agent wanted
lt--gt patient
54
Challenge 5 Omissible constituents

• Intuition
• Same context, two expressions that differ by one
  constituent ? a general construction with the
  constituent being omissible
• May require verbatim memory traces of utterances
  relevant context

55
When does the learning stop?
Bayesian Learning Framework

• Most likely grammar given utterances and context
• The grammar prior is a preference for the kind
  of grammar
• In practice, take the log and minimize cost ?
  Minimum Description Length (MDL)

56
Intuition for MDL

• S -gt Give me NP
• NP -gt the book
• NP -gt a book

• S -gt Give me NP
• NP -gt DET book
• DET -gt the
• DET -gt a

Suppose that the prior is inversely proportional
to the size of the grammar (e.g. number of
rules) Its not worthwhile to make this
generalization
57
Intuition for MDL

• S -gt Give me NP
• NP -gt the book
• NP -gt a book
• NP -gt the pen
• NP -gt a pen
• NP -gt the pencil
• NP -gt a pencil
• NP -gt the marker
• NP -gt a marker

• S -gt Give me NP
• NP -gt DET N
• DET -gt the
• DET -gt a
• N -gt book
• N -gt pen
• N -gt pencil
• N -gt marker

58
How to calculate the prior of this grammar

• (Yet to be determined)
• There is evidence that the lexicalized
  constructions do not completely go away
• If the more lexicalized constructions are
  retained, the size of grammar is a bad indication
  of degree of generality