AI Programming Lecture 7 Natural Language Processing

1
AI ProgrammingLecture 7Natural Language
Processing

• Richard Price
• School Of Computer Science University Of
  Birmingham
• msc59rmp_at_cs.bham.ac.uk

2
Last Week

• Sub-goals.
• Planning.
• STRIPS

3
Introduction

• Definition.
• Applications.
• Difficulties.
• Syntax Grammar.
• Lexicon.
• Semantics.

4
Definition

• Natural Language
• Spoken.
• Written.
• Program languages?
• Words change quickly over time.
• Language structure slowly changes.

5
Applications

• Databases.
• Automated booking systems.
• Machine Translation.
• Operating Systems.
• Summarising documents.
• Others?

6
Difficulties

• How do we
• Specify requirements of a system.
• Receive language input.
• Understand language.
• Respond intelligently.

7
Syntax

• The syntax is the structure of a phrase.
• A grammar is the rules within a language.
• Example Grammar
• S ? NP VP
• NP ? det noun
• VP ? verb NP

8
Lexicon

• The lexicon is the set of rules of individual
  rules.
• Example
• verb ? saw
• verb ? like
• noun ? man
• noun ? woman
• det ? a
• det ? the
• What about email?

9
Syntax Trees
S
NP
VP
det
noun
NP
v
the
man
saw
det
noun
the
woman
10
Top Down Parser

• For actual code see Sheilas notes.
• backwardsParseTree()
• Start with high level goal structure e.g S.
• Taking the rules in order
• If the LHS of a rule matches our goal or
  sub-goal.
• Then explore the RHS of this rule as a sub-goal.
• Until either
• We complete all our goals sub-goals.
• We run out of rules.
• Return structure.

11
Left Recursive Rules

• Complex sentences require more rules.
• The big red computer.
• Add to existing grammar
• NP ? det AP noun
• AP ? AP adj
• AP ? adj
• adj ? big
• adj ? red
• adj ? stupid
• What happens now?

12
Bottom Up Recogniser

• forwardsParseGoals()
• Start with low level input(s) as goal.
• Taking the rules in order
• If the RHS of a rule matches our goal or
  sub-goal.
• Then replace our goal or sub-goal with LHS of
  rule.
• Until either
• We finish with a single goal matching our
  structure.
• We run out of rules.
• Return True or False.

13
Ambiguity

• Many words fall under several categories.
• Email, fax, fly
• This causes ambiguity.
• Structures can also be ambiguous.
• I saw her email.
• Ambiguous structures may indicate ambiguous
  meanings.