CS 224U LINGUIST 288188 Natural Language Understanding Jurafsky and Manning

1
CS 224U / LINGUIST 288/188Natural Language
UnderstandingJurafsky and Manning

• Semantic/Thematic Roles
• Oct 9, 2007
• Christopher Manning

2
Semantic roles in public life
3
Semantically precise search for relations/events
Query afghans destroying opium poppies
4
What are semantic roles and what is their
history?

• A lot of forms of traditional grammar (Sanskrit,
  Japanese, ) analyze in terms of a rich array of
  semantically potent case ending or particles
• Theyre kind of like semantic roles
• The idea resurfaces in modern generative grammar
  in work of Charles (Chuck) Fillmore, who calls
  them Case Roles (Fillmore, 1968, The Case for
  Case).
• Theyre quickly renamed to other words, but
  various
• Semantic roles
• Thematic roles
• Theta roles
• A predicate and its semantic roles are often
  taken together as an argument structure

5
Okay, but what are they?

• An event is expressed by a predicate and various
  other dependents
• The claim of a theory of semantic roles is that
  these other dependents can be usefully classified
  into a small set of semantically contentful
  classes
• And that these classes are useful for explaining
  lots of things

6
Common semantic roles

• Agent initiator or doer in the event
• Patient affected entity in the event undergoes
  the action
• Sue killed the rat.
• Theme object in the event undergoing a change of
  state or location, or of which location is
  predicated
• The ice melted
• Experiencer feels or perceive the event
• Bill likes pizza.
• Stimulus the thing that is felt or perceived

7
Common semantic roles

• Goal
• Bill ran to Copley Square.
• Recipient (may or may not be distinguished from
  Goal)
• Bill gave the book to Mary.
• Benefactive (may be grouped with Recipient)
• Bill cooked dinner for Mary.
• Source
• Bill took a pencil from the pile.
• Instrument
• Bill ate the burrito with a plastic spork.
• Location
• Bill sits under the tree on Wednesdays

8
Common semantic roles

• Try for yourself!
• The submarine sank a troop ship.
• Doris hid the money in the flowerpot.
• Emma noticed the stain.
• We crossed the street.
• The boys climbed the wall.
• The chef cooked a great meal.
• The computer pinpointed the error.
• A mad bull damaged the fence on Jacks farm.
• The company wrote me a letter.
• Jack opened the lock with a paper clip.

9
Okay, but why are they useful?

• Linguistics
• Syntax and morphology often reflects semantic
  roles
• E.g., in many Indian languages experiencer
  subjects appear in the dative not the
  nominative/ergative case
• E.g., in English theme but not agent subjects
  allow locative inversion
• Among the guests was sitting my friend Rose
• Among the guests was talking my friend Rose
• Towards me lurched a drunk
• Towards me yelled a drunk
• Human conceptualization of the world
• Viewing events in terms of actors and undergoers
  is not inherently part of the world its the
  human conceptualization (categorization) of the
  world
• Cf. also figure/ground theme/location

10
Diathesis Alternations

• Alternations
• Spray / Load
• Hit / Break
• Non-alternating
• Swat / Dash
• Fill / Cover

Slides from Martha Palmer
11
Hit / Break Alternation

• John hit the fence with a stick.
• John hit the stick against a fence.
• John broke the fence with a stick.
• John broke the stick against the fence.
• Radical change in meaning associated with break
  but not hit.
• Can be explained via semantic roles or
  proto-roles (change of state for direct object
  with break class).

12
Problems with Thematic Role Types

• Fragmentation Cruse (1973) subdivides Agent
  into four types.
• Ambiguity Andrews (1985) is Instrument, an
  adjunct or a core argument?
• Symmetric stative predicates e.g. This is
  similar to that Distinct roles or not?
• Searching for a Generalization What is a
  Thematic Role?

Slides from Michael Mulyar
13
Proto-Roles (Dowty 1991)

• Event-dependent Proto-roles introduced
• Prototypes based on shared entailments
• Grammatical relations such as subject related to
  observed (empirical) classification of
  participants
• Typology of grammatical relations
• Proto-Agent
• Proto-Patient

14
Proto-Agent and Proto-Patient

• Proto-Agent Properties
• Volitional involvement in event or state
• Sentience (and/or perception)
• Causing an event or change of state in another
  participant
• Movement (relative to position of another
  participant)
• (exists independently of event named)
• may be discourse pragmatic
• Proto-Patient Properties
• Undergoes change of state
• Incremental theme
• Causally affected by another participant
• Stationary relative to movement of another
  participant
• (does not exist independently of the event, or at
  all) may be discourse pragmatic

15
The rebirth of computational work on semantic
roles

• Various kinds of early conceptual or semantic
  parsing work had semantic roles
• Schank had something like semantic roles
• Rebirth was by reconceptualizing semantic roles
  as a classification task following parsing
• Daniel Gildea, Daniel Jurafsky Automatic
  Labeling of Semantic Roles. ACL 2000
• Daniel Gildea, Daniel Jurafsky Automatic
  Labeling of Semantic Roles. Computational
  Linguistics 28(3) 245-288 (2002)

16
What enabled this work?

• Labeled training data (supervised learning!!)
• A tale of two resources
• FrameNet
• Fillmore, Baker, Wooters, Lowe, Jurafsky
• Founded on lexicographic principles
• Pays much more attention to the connection
  between language and conceptualization
• PropBank
• Martha Palmer
• Founded on linguistic principles (Dowty, Levin)
• Much more designed to be machine learning data
• Provides coverage at the expense of depth

17
Whats FrameNet?

• The FrameNet lexical database, currently contains
  more than 10,000 lexical units, more than 6,100
  of which are fully annotated, in more than 825
  semantic frames, exemplified in more than 135,000
  annotated sentences.
• Has properties of Dictionary and Thesaurus
• Unlike WordNet FN shows word-to-word
  relationships via semantic frames.
• Unlike COMLEX FN links syntactic patterns with
  semantic patterns.
• Unlike MindNet based on manual operations on
  corpus data

These slides from Charles Fillmore
18
Basic Assumptions

• Word meanings are best understood in terms of the
  semantic/conceptual structures which they
  presuppose. (We use the word frame
  promiscuously to cover all such structures, big
  and small.)
• A frame has various elements and words (as well
  as grammatical constructions) are said to evoke
  frames and their elements.

19
BARTENDER

• The word bartender
• evokes a frame of the preparation and serving
  of alcoholic beverages, and
• profiles the individual who does this work.
  The bartender asked me for my ID.

20
The Work of FrameNet

• To discover and describe those frames that
  support lexical meanings,
• to provide names for the relevant elements of
  those frames,
• to describe the syntactic/semantic valence of the
  words that fit the frames, and
• to base the whole process on attestations from a
  corpus.

21
The Beginning

• Fillmore and Atkins first proposed developing a
  frame-based lexicon, and exemplified it in
  studies of the English word risk, as verb and
  noun.
• One aspect of the polysemy of risky situations
  can be represented in a diagram inspired by work
  in decision theory circle Chance, square
  Choice.

22
Frame Elements Risk PR protagonist BA bad
outcome DE deed PO possession GO goal
DE
BA
DE
GO
23
(Heavily Edited Examples)

• He does not want to risk a rebuff.
• He had risked his reputation.
• He had to risk Pop getting mad at him.
• He risked his life.
• He risks committing a grave mistake.
• I couldnt risk leaving my vantage point.
• I risked a pause to let that sink in.
• I wouldnt risk letting you meet her.
• We decided to risk the venture.
• She risked going into the pool alone.
• You would risk death doing what she did.

24
Phrase Types
He does not want to risk a rebuff. He had risked
his reputation. He had to risk Pop getting mad at
him. He risked his life. He risks committing a
grave mistake. I couldnt risk leaving my vantage
point. I risked a pause to let that sink in. I
wouldnt risk letting you meet her. We decided to
risk the venture. She risked going into the pool
alone. You would risk death doing what she did.
25
Observed PTs for Objects

• NP death, his reputation
• VP Gerund letting you rest here
• S Gerund Pop getting made at you

26
(No Transcript)
27
Semantic Roles (FEs)

• BA (bad outcome) death, Pop getting mad at you
• DE (deed, decision leading to risk) the venture,
  leaving my vantage point
• PO (valued possession, asset) his reputation,
  your life

28
Frame Elements and Phrase Types
BA
NP
Gerund
death bankruptcy
losing your job Pop getting mad at you
DE
Gerund
NP
a trip into the jungle the venture
swimming here speaking up
PO
NP
my reputation your inheritance
29
All this for one word??

• Other words participate in the same background
  frame as well jeopardize chance hazard g
  amble bet wager

30
Frame-Related Words

• The real interest of FrameNet (in fact) has been
  in setting up frames that are needed for the
  description of large families of words, and also
  frames that participate in families of derivative
  frames.
• (The Net idea in FrameNet has to do with
  linking words to frames, and frames to other
  frames.)

31
(No Transcript)
32
PropBank starts from a Penn TreeBank phrase
structure
a GM-Jaguar pact that would give the U.S. car
maker an eventual 30 stake in the British
company.
NP
SBAR
S
WHNP-1
VP
that
NP-SBJ
VP
T-1
would
NP
give
PP-LOC
Slides from Martha Palmer
33
The same phrase, PropBanked
a GM-Jaguar pact that would give the U.S. car
maker an eventual 30 stake in the British
company.
Arg0
that would give
Arg1
T-1
an eventual 30 stake in the British company
Arg2
the US car maker
34
The full sentence, PropBanked
have been expecting
Arg1
Arg0
Analysts have been expecting a GM-Jaguar pact
that would give the U.S. car maker an eventual
30 stake in the British company.
Analysts
Arg0
that would give
Arg1
T-1
an eventual 30 stake in the British company
Arg2
the US car maker
35
PropBank

• 1M words of WSJ annotated with predicate-argument
  structures for verbs.
• The location type of each verbs arguments
• Argument types are defined on a per-verb basis.
• Consistent across uses of a single verb (sense)
• But the same tags are used (Arg0, Arg1, Arg2, )
• Arg0 ? proto-typical agent (Dowty)
• Arg1 ? proto-typical patient

Next 6 slides from Ed Loper
36
PropBank frame files cover (smear, put over)

• Arguments
• Arg0 causer of covering
• Arg1 thing covered
• Arg2 covered with
• Example
• John covered the bread with peanut butter.

37
PropBank Trends in Argument Numbering

• Arg0 proto-typical agent (Dowty)
• Agent (85), Experiencer (7), Theme (2),
• Arg1 proto-typical patient (Dowty)
• Theme (47),Topic (23), Patient (11),
• Arg2 Recipient (22), Extent (15), Predicate
  (14),
• Arg3 Asset (33), Theme2 (14), Recipient
  (13),
• Arg4 Location (89), Beneficiary (5),
• Arg5 Location (94), Destination (6)

38
PropBank Adjunct Tags

• Variety of ArgMs (Arggt5)
• TMP when?
• LOC where at?
• DIR where to?
• MNR how?
• PRP why?
• REC himself, themselves, each other
• PRD this argument refers to or modifies another
• ADV others

39
Limitations to PropBank as Training Data

• Args2-5 seriously overloaded ? poor performance
• FrameNet (and VerbNet) both provide more
  fine-grained role labels
• Example
• Rudolph Agnew,, was named ARG2/Predicate a
  nonexecutive director of this British industrial
  conglomerate.
• .the latest results appear in todays New
  England Journal of Medicine, a forum likely to
  bring new attention ARG2/Destination to the
  problem.
• WSJ too domain specific too financial.
• Need broader coverage genres for more general
  annotation.

40
PropBank/FrameNet
Buy Arg0 buyer Arg1 goods Arg2
seller Arg3 rate Arg4 payment
Sell Arg0 seller Arg1 goods Arg2
buyer Arg3 rate Arg4 payment