Artificial Intelligence Chapter 22: Communication

1
Artificial IntelligenceChapter 22 Communication

• Michael Scherger
• Department of Computer Science
• Kent State University

2
Contents

• Spoken Language
• Written Language
• Grammars
• Language Understanding
• Language Generation

3
Introduction

• We know how to work with predicate calculus
• (On Box x) (Color x Red)
• Resolution, STRIPS, CLIPS, etc
• How do you get predicate calculus from natural
  language?

4
Introduction

• Must have a representation of the environment in
  which the sentence sequence is understood
• THE WORLD
• The world (for robots) consists of
• Objects
• Relations between objects
• Properties of objects
• Actions

5
Introduction
Language
Spoken Language
Written Language
Isolated Words
Connected Speech
Constrained Vocabulary
Unconstrained Vocabulary
Constrained Vocabulary
Unconstrained Vocabulary
6
Spoken Language

• Speech Recognition
• Translate spoken language to words
• acoustical pattern recognition
• Phoneme fundamental speech sound
• English has app. 40 phonemes
• Syllable combination of phonemes
• English has app. 10,000unique syllables

7
Spoken Language

• Speech Understanding
• Represent meaning of sequence of spoken words
• Combines Speech Recognition and Written Language
  Understanding

8
Speech Recognition

• Isolated Word Recognition
• Words are separated by pauses
• May filter out background noise
• Performs pattern matching on digitized sound
• Speaker dependent systems
• May have higher success rates
• May have larger vocabularies
• Require training for individual users
• Speaker independent systems
• Have smaller vocabularies (20 1000 words)
• Do not require training

9
Speech Recognition

• Connected Speech (Continuous) Recognition
• No forced pauses between words
• Like natural human communication
• More difficult than written language
  understanding
• Finding individual words can be difficult
• Homophones (too two)
• Phonetic ambiguity (I scream ice cream)
• Syllable omission (govment government)
• Ungrammatical speech

10
Speech Synthesis

• Translate a text representation of a word to
  speech
• Break word up into corresponding phonemes
• Largely an algorithmic problem (not considered
  AI)
• Synthesized speech is commercially available and
  part of some OSes
• Adequate performance
• http//www.microsoft.com/msagent/downloads/user.as
  p
• http//www.mheller.com/agent.htm

11
Written Language

• Focus on machine-readable representation of words
• NOT on character recognition
• Must examine deep structure not just surface
  syntax
• Understanding requires syntax and semantics
• John is easy to please
• John is eager to please
• Conceptual Dependency (CD) is one way of
  representing deep structure
• But first we need to start with surface syntax

12
Real Language

• Real human languages provide many problems for
  NLP
• ambiguity
• anaphora
• indexicality
• vagueness
• noncompositionality
• discourse structure
• metonymy
• metaphor

13
Ambiguity

• Squad helps dog bite victim
• Helicopter powered by human flies
• American pushes bottle up Germans
• I ate spaghetti with meatballs
• salad
• abandon
• a fork
• a friend
• Ambiguity can be lexical (polysemy), syntactic,
  semantic, referential

14
Indexicality

• Indexical sentences refer to utterance situation
  (place, time, S/H, etc.)
• I am over here
• Why did you do that?

15
Anaphora

• Using pronouns to refer back to entities already
  introduced in the text
• After Mary proposed to John, they found a
  preacher and got married.
• For the honeymoon, they went to Hawaii
• Mary saw a ring through the window and asked John
  for it
• Mary threw a rock at the window and broke it
• Associative Calculus

16
Metonymy

• Using one noun phrase to stand for another
• I've read Shakespeare
• Chrysler announced record profits
• The ham sandwich on Table 4 wants another beer

17
Metaphor

• Non-literal" usage of words and phrases, often
  systematic
• I've tried killing the process but it won't die.
  Its parent keeps it alive.

18
Noncompositionality

• basketball shoes
• baby shoes
• alligator shoes
• designer shoes
• brake shoes
• red book
• red pen
• red hair
• red herring
• small moon
• large molecule
• mere child
• alleged murderer
• real leather
• artificial grass

19
Grammars

• Focus on machine-readable representation of words
• NOT on character recognition
• Must examine deep structure not just surface
  syntax
• Understanding requires syntax and semantics
• John is easy to please
• John is eager to please
• Conceptual Dependency (CD) is one way of
  representing deep structure
• But first we need to start with surface syntax

• Grammar
• T
• Finite set of terminal symbols
• V
• Finite set of variables or non-terminal symbols
• s
• Sentence to be constructed
• P
• Set of production rules
• How V and T can be combined

20
Grammar Example

• T 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
• Digits
• s
• Number
• P
• Number is single digit
• Number is single digit followed by a number

• Grammar
• T
• Finite set of terminal symbols
• V
• Finite set of variables or non-terminal symbols
• s
• Sentence to be constructed
• P
• Set of production rules
• How V and T can be combined

21
BNF Grammar Example

• T
• 0, 1, 2, 3, 4, 5, 6, 7, 8, 9
• V
• ltnumbergt
• s
• ltnumbergt
• P
• ltnumbergt ltdigitgt ltdigitgtltnumbergt
• ltdigitgt 0123456789

• Grammar
• T
• Finite set of terminal symbols
• V
• Finite set of variables or non-terminal symbols
• s
• Sentence to be constructed
• P
• Set of production rules
• How V and T can be combined

22
Wumpus Lexicon
23
Wumpus Grammar
24
Wumpus Parse Tree
25
Language Understanding

• Accept sequence of words as input
• Parse sequence
• Determine syntactic category of words
• See if sequence is a sentence
• Create and store an internal representation
• Analyze sentence
• Semantic analysis
• Type of sentence
• Declarative
• Store info in knowledge base
• Interrogative
• Retrieve info from knowledge base
• Imperative
• Execute corresponding action

26
Phase Structure Grammar

• ltsentencegt ltnoun phrasegt ltverb phrasegt
• ltnoun-phrasegt ltnoungt ltadjgt ltnoungt
  ltdeterminergt ltnoungt
• ltverb-phrasegt ltverbgt ltverbgt ltnoun-phrasegt
• ltdeterminergt A THE
• ltverbgt IS IS-A IS-KNOWN SUPPORTS
• ltadjgt RIGHT-OF LEFT-OF LARGE SMALL
  etc
• ltnoungt SHAPE COLOR SIZE BLOCK
  PYRAMID etc

27
Parse Tree

• WHAT BLOCK IS SUPPORTED-BY B5

• ltsentencegt ltnoun phrasegt ltverb phrasegt
• ltnoun-phrasegt ltnoungt ltadjgt ltnoungt
  ltdeterminergt ltnoungt
• ltverb-phrasegt ltverbgt ltverbgt ltnoun-phrasegt
• ltdeterminergt A THE
• ltverbgt IS IS-A IS-KNOWN SUPPORTS
• ltadjgt RIGHT-OF LEFT-OF LARGE SMALL
  etc
• ltnoungt SHAPE COLOR SIZE BLOCK
  PYRAMID etc

28
Drawback for Simple Binary Tree Parser

• Ignored plural forms
• Could convert atoms as strings and add s or
  es
• But what about complex plurals
• E.g. person people, mouse mice, goose geese
• No prefixes or suffixes
• convert atoms to strings
• Each word had only one part of speech
• Many words have multiple parts of speech
• Store (n) and store (v), hit (adj) and hit (v)

29
Transition Network

• Semantic Network
• Used to represent more complex grammars
• Nodes
• Named state
• Arcs
• Named transition from one node to another
• Parsing a transition network
• Traversing the network based on an input phrase
  or sentence

30
Arcs or Transitions

• Simple Transition Network (STN)
• CAT
• Matches word with a category
• CAT noun
• WORD
• Matches an individual word
• WORD study
• JUMP
• Go directly to next node
• POP
• Exit network
• Recursive Transition Network (RTN) and Augmented
  Transition Network (ATN)
• PUSH
• Transfer control to another network
• PUSH verb-phrase

31
STN Example

• Noun Phrase Transition Network
• Start at NP0
• Exit at NP1
• CATs match the POS categories from grammar
  definition
• Angle brackets not typically used in TNs
• Test matches according to number

32
STN Example

• Check Phrases
• The shape
• Large block
• The large block
• B1

33
STN Example
THE BLOCK FELL
34
More Ambiguity

• Words may have multiple meaning
• Computer must determine which meaning to use
  (disambiguation)
• Beth waved her hand
• Larry asked Joan for her hand
• Cathy gave Amy a hand with her project
• Dean is an old hand at fixing things
• Time flies like an arrow
• Fruit flies like a banana
• The vodka is good but the meat is rotten
• The spirit is willing but the flesh is weak

35
More Ambiguity

• It may be impossible to make correct sense of one
  statement without having access to and correct
  interpretations of prior sentences.
• Harry and Clarence are brothers. Harry is older.
• Kathy went to the mall to buy a gift. She bought
  it.

36
Pragmatics

• In spoken language, simple statements or
  questions may in fact contain implicit requests
  or instructions.
• Could you pass the bread?
• Do you know the way to San Jose?

37
Understanding!!!!!

• Understanding is the conversion of input into a
  form that allows
• Questions to be answered
• Plans to be made
• Decisions/Actions to be carried out
• Semantic Analysis is frequently based on case

38
Understanding

• Case analysis determines the role of noun phrases
  (i.e. objects)
• In many European languages, case can be
  determined by endings and other clues
• In Englishuse
• prepositions
• verbs
• position

39
Understanding

• Most common cases are
• Agent
• Beneficiary
• Instrument
• Conveyance
• Location
• Object
• Co-agent
• Time
• Source
• Destination

40
Understanding

• Agent
• Causes the action to occur
• Position
• Usually first noun phrase without a preposition
• Subject of the sentence
• Ex The car hit the wall
• Beneficiary
• For whom something is done
• Preposition for
• Ex John did it for Mary
• Possible restrictionmust be for people

41
Understanding

• Instrument
• Means of accomplishing the action
• Preposition with
• Ex John hit it with a book
• Possible restrictionmust not be people

42
Understanding

• Conveyance
• Means of movement
• Preposition by
• Ex John went to town by train
• Possible restrictionmust not be people

43
Understanding

• Location
• Where the action takes place
• Preposition on, by, in
• Ex John sat by the window
• Object
• The thing undergoing the state change or
  receiving the action
• Position
• The noun phrase after the verb without a
  preposition
• CautionMore complex analysis is necessary!!!!!
• Ex John saw the book

44
Understanding

• Co-Agent
• Preposition with
• Ex John went with Joan to the store
• Possible restrictionmust be people

45
Understanding

• Source and Destination
• Prepositions from and to
• Ex John went from home to work
• Ex It went from bad to worse
• Trajectory
• The motion from source to destination
• Ex They went in through the door
• Ex He carried her over the threshold

46
Understanding

• Note that questions can be answered by case
• What is answered by object
• Who is answered by agent
• With what is answered by instrument
• Whombeneficiary