CPSC 503 Computational Linguistics

1
CPSC 503Computational Linguistics

• Semantic Analysis
• Lecture 16
• Giuseppe Carenini

2
Semantic Analysis
Sentence
Meanings of grammatical structures

• Syntax-driven
• Semantic Analysis

Meanings of words
Literal Meaning
I N F E R E N C E
Common-Sense Domain knowledge
Further Analysis
Discourse Structure
Intended meaning
Context
3
Today 17/3

• Compositional Analysis
• Integrate semantics and parsing
• Non-compositionality
• Semantic Grammars
• Information Extraction

4
Meaning Structure of Language

• How does language convey meaning?
• Grammaticization
• Tense systems
• Conjunctions
• Quantifiers
• Indefinites (variables)
• Display a partially compositional semantics
• Display a basic predicate-argument structure

5
Compositional Analysis

• Principle of Compositionality
• The meaning of a whole is derived from the
  meanings of the parts
• What parts?
• The constituents of the syntactic parse of the
  input
• What could it mean for a part to have a meaning?

6
Compositional Analysis Example

• AyCaramba serves meat

7
Augmented Rules

• Augment each syntactic CFG rule with a semantic
  formation rule

• i.e., The semantics of A can be computed from
  some function applied to the semantics of As
  parts.

• The class of actions performed by f will be quite
  restricted.

8
Simple Extension of FOL Lambda Forms

• Lambda-reduction variables are bound by treating
  the lambda form as a function with formal
  arguments

9
Augmented Rules Example
assigning constants

• Easy parts

• Attachments
• AyCaramba
• MEAT

• PropNoun -gt AyCaramba
• MassNoun -gt meat

10
Augmented Rules Example
Semantics attached to one daughter is applied to
semantics of the other daughter(s).

• VP.sem(NP.sem)
• Verb.sem(NP.sem)

• S -gt NP VP
• VP -gt Verb NP

lambda-form

• Verb -gt serves

11
Example
MEAT
.
AC
MEAT

• S -gt NP VP
• VP -gt Verb NP
• Verb -gt serves
• NP -gt PropNoun
• NP -gt MassNoun
• PropNoun -gt AyCaramba
• MassNoun -gt meat

• VP.sem(NP.sem)
• Verb.sem(NP.sem)
• PropNoun.sem
• MassNoun.sem
• AC
• MEAT

12
Problem Quantified Phrases

• Consider
• A restaurant serves meat.
• Assume that semantics for A restaurant is

• If we proceed as we did in the previous example,
  the semantics for S would be

?!?
13
Solution Complex Terms
Complex-Term ? ltQuantifier var bodygt
Examples
14
Convert complex-terms back to FOL
P(ltquantifier, var, bodygt)
Quantifier var body connective P(var)
Example
15
Problem Quantifier Scope Ambiguity

• Consider Every restaurant has a menu

16
Solution Quantifier Scope Ambiguity

• Similarly to PP attachment, number of possible
  interpretations exponential in the number of
  complex terms

17
Attachments for a fragment of English

• Sentences
• Noun-phrases
• Verb-phrases
• Prepositional-phrases

Based on The core Language Engine 1992
18
Integration with a Parser

• Assume youre using a dynamic-programming style
  parser (Earley or CYK).

• Two basic approaches
• Integrate semantic analysis into the parser
  (assign meaning representations as constituents
  are completed)

• Pipeline assign meaning representations to
  complete trees only after theyre completed

19
Pros and Cons

• Integration
• use semantic constraints to cut off parses that
  make no sense
• assign meaning representations to constituents
  that dont take part in the correct (most
  probable) parse

20
Non-Compositionality

• Unfortunately, there are lots of examples where
  the meaning of a constituent cant be derived
  from the meanings of the parts

• - metaphor, (corporation as person)
• metonymy, (??)
• idioms,
• irony,
• sarcasm,
• indirect requests, etc

21
English Idioms

• Lots of these constructions where the meaning of
  the whole is either
• Totally unrelated to the meanings of the parts
  (kick the bucket)
• Related in some opaque way (run the show)

• buy the farm
• bite the bullet
• bury the hatchet
• etc

22
The Tip of the Iceberg

• Enron is the tip of the iceberg.
• NP -gt the tip of the iceberg .

• the tip of an old iceberg
• the tip of a 1000-page iceberg
• the merest tip of the iceberg

NP -gt TipNP of IcebergNP TipNP NP with tip
as its head IcebergNP NP with iceberg as its
head
23
Handling Idioms

• Mixing lexical items and grammatical constituents
• Introduction of idiom-specific constituents
• Permit semantic attachments that introduce
  predicates unrelated with constituents

NP -gt TipNP of IcebergNP small-part(),
beginning(). TipNP NP with tip as its head
IcebergNP NP with iceberg as its head
24
Knowledge-Formalisms Map(including probabilistic
formalisms)
State Machines (and prob. versions) (Finite State
Automata,Finite State Transducers, Markov Models)
Morphology
Syntax
IE
Rule systems (and prob. versions) (e.g., (Prob.)
Context-Free Grammars)
SG
Semantics

• Logical formalisms
• (First-Order Logics)

Pragmatics Discourse and Dialogue
AI planners
25
Semantic Grammars

• Def CFGs in which rules and constituents
  correspond directly to semantic entities and
  relations

26
Semantic Grammars

• Limitations
• Almost complete lack of reuse
• Tend to grow in size (missing syntactic
  generalizations)

• Typically used in conversational agents in
  constrained domains
• Limited vocabulary
• Limited grammatical complexity

27
Information Extraction (IE)

• Scanning newspapers, newswires for a fixed set of
  events of interests E.g., ??
• Scanning websites for products, prices, reviews,
  etc.

• Arbitrarily complex (long) sentences
• Extended discourse
• Multiple writers

28
Back to Finite State Methods

• Apply a series of cascaded transducers to an
  input text

• At each stage specific elements of
  syntax/semantics are extracted for use in the
  next level e.g., complex phrases, semantic
  patterns

• The end result is a set of relations suitable for
  entry into a database

29
Complex Phrases Semantic Patterns

• Bridgestone Sports Co. said Friday it has set up
  a joint venture in Taiwan with a local concern
  and a Japanese trading house to produce golf
  clubs to be shipped to Japan.
• The joint venture, Bridgestone Sports Taiwan
  Co., capitalized at 20 million new Taiwan
  dollars, will start production in January 1990
  with production of 20,000 iron and metal wood
  clubs a...

30
FASTUS Output
31
Named Entities Recognition

• Labeling all the occurrences of named entities in
  a text
• People, organizations, lakes, bridges, hospitals,
  mountains, etc

• This can be done quite robustly and looks like
  one of the most useful tasks across a variety of
  applications

32
Next Time

• Lexical Semantics
• Read Chapter 16

33
Meaning Structure of Language

• The semantics of human languages
• Display a basic predicate-argument structure
• Make use of variables
• Make use of quantifiers
• Use a partially compositional semantics

34
IE Key Points

• What about the stuff we dont care about?
• Ignore it. I.e. Its not written to the next
  tape, so it just disappears from further
  processing

• It works because of the constrained nature of the
  problem
• Only looking for a small set of items that can
  appear in a small set of roles

35
Cascades
36
Key Point

• It works because of the constrained nature of the
  problem
• Only looking for a small set of items that can
  appear in a small set of roles

37
Next Time

• More robust approaches to semantic analysis
• Semantic grammars
• Information extraction
• Probabilistic labeling
• More on less than compositional constructions and
• Word meanings
• So read Chapter 16

38
Predicate-Argument Semantics

• The functions/operations permitted in the
  semantic rules fall into two classes
• Pass the semantics of a daughter up unchanged to
  the mother
• Apply (as a function) the semantics of one of the
  daughters of a node to the semantics of the other
  daughters

39
Predicate-Argument Semantics

• S -gt NP VP
• VP -gt Verb NP
• Is it really necessary to specify these
  attachments?

• VP.sem(NP.sem)
• Verb.sem(NP.sem)
• No, in each rule theres a daughter whose
  semantics is a function and one that isnt. What
  else is there to do?

40
Harder Example

• What makes this hard?
• What role does Harry play in all this?

41
Harder Example

• The VP for told is VP -gt V NP VPto
• So you do what?
• Apply the semantic function attached to VPTO the
  semantics of the NP this binds Harry as the goer
  of the going.
• Then apply the semantics of the V to the
  semantics of the NP this binds Harry as the
  Tellee of the Telling
• And to the result of the first application to
  get the right value of the told thing.
• V.Sem(NP.Sem, VPto.Sem(NP.Sem)

42
Harder Example

• Thats a little messy and violates the notion
  that the grammar ought not to know much about
  what is going on in the semantics
• Better might be
• V.sem(NP.Sem, VPto.Sem)
• i.e Apply the semantics of the head verb to the
  semantics of its arguments.
• Complicate the semantics of the verb inside VPto
  to figure out whats going on.

43
Two Philosophies

• Let the syntax do what syntax does well and dont
  expect it to know much about meaning
• In this approach, the lexical entrys semantic
  attachments do the work
• Assume the syntax does know about meaning
• Here the grammar gets complicated and the lexicon
  simpler

44
Example

• Consider the attachments for the VPs
• VP -gt Verb NP NP rule (gave Mary a book)
• VP -gt Verb NP PP (gave a book to Mary)
• Assume the meaning representations should be the
  same for both. Under the lexicon-heavy scheme the
  attachments are
• VP.Sem(NP.Sem, NP.Sem)
• VP.Sem(NP.Sem, PP.Sem)

45
Example

• Under the syntax-heavy scheme we might want to do
  something like
• VP -gt V NP NP
• V.sem Recip(NP1.sem)
  Object(NP2.sem)
• VP -gt V NP PP
• V.Sem Recip(PP.Sem) Object(NP1.sem)
• I.e the verb only contributes the predicate, the
  grammar knows the roles.

46
Constructional Approach

• So well allow both
• VP ? V NP V.sem(NP.sem)
• and
• VP ? Kick-Verb the bucket ? x Die(x)

47
Semantic Grammars

• One problem with traditional grammars is that
  they dont necessarily reflect the semantics in a
  straightforward way
• You can deal with this by
• Fighting with the grammar
• Complex lambdas and complex terms, etc
• Rewriting the grammar to reflect the semantics
• And in the process give up on some syntactic
  niceties

48
BERP Example
49
BERP Example

• How about a rule like the following
• Request ? I want to go to eat FoodType Time
• some attachment

50
Semantic Grammar

• The term semantic grammar refers to the
  motivation for the grammar rules
• The technology (plain CFG rules with a set of
  terminals) is the same as weve been using
• The good thing about them is that you get exactly
  the semantic rules you need
• The bad thing is that you need to develop a new
  grammar for each new domain

51
Semantic Grammars

• Typically used in conversational agents in
  constrained domains
• Limited vocabulary
• Limited grammatical complexity
• Chart parsing (Earley) can often produce all
  thats needed for semantic interpretation even in
  the face of ungrammatical input.