Introduction to Knowledge Representation

Two Perspectives

• Artificial Intelligence
• All intelligent systems exhibit (complex) behavior
• Flexibility of behavior
• Language and communication
• We analyze intelligent systems
• in terms of beliefs
• We influence intelligent systems
• by influencing beliefs (e.g., talking)
• Information systems
• E.g., database applications
• CMS, inventory, POS, medical records
• Characterized by manipulating large datasets
• Retrieval is key
• Think libraries!

Information Systems

• Some programs are information heavy
• Address books, mail programs
• Some are not
• Compilers, word processors, etc.
• Ideally, separate the information management from the rest
• Enhance maintenance, reuse, etc.
• Can't always get a clean separation

Regex Analogy (1)

• Consider simple language for regular expressions
• a-z, ! 
• * for zero or more
• | for alternation
• (Parens for precedence)
• We can express all sorts of fun things:
• aa* (one or more as)
• Gooo*gle (the Google search bar)
• Yahooo*!|Gooo*gle (If Yahoo adopted the convention)
• (Ya|G)ooo*(gle|!)
• Standard services:
• match("(Ya|G)ooo*(gle|!)", "Gogle") ⇒  false
• extract("Yahooo*!","<b>Yahoo!</b>") ⇒ "Yahoo!"
• replace("aaa*","a","Aaannoyance") ⇒ "Annoyance"

Regex Analogy (2)

• Non-standard services
• equiv("(Ya|G)ooo*(gle|!)","Yahooo*!|Gooo*gle")
• contained_in("Yahooo*!","Ya|G)ooo*(gle|!)")
• Consider coding alternatives:

• if match("Yahooo*!","Gooogle") or \ 
   match("Gooo*gle","Gooogle"):
   print "A search results page!"
  
  if match("Yahooo*!|Gooo*gle","Gooogle"):
   print "A search results page!"

• The first hides the overall structure from people and the regex engine
• Can't transform or optimize
• Hard to reuse (e.g., in other langauges)

Knowledge Representation

• IS related definition
• Rich, declarative representation of information
• Declarative
• Specifies the What, not the How
• Rich
• Can express complex information
• To people and programs
• Can analyze representations
• Can derive implicit information
• Two key aspects
• Representation (e.g., expressivity, what can you say and how)
• Reasoning (what can you do with the representation?)

3 Foci of KR

• Represent commonsense knowledge
• Cyc project 
• Naive physics mainfesto
• Represent scientific knowledge
• A branch of scientific computing
• Big in bio-informatics
• Other specialized areas: law, medical-informatics, etc.
• Conceputal modeling
• UML, CRC cards, ER diagrams
• Related to Formal Methods
• These do not differ so much in domain but in style
• also, purpose, detail, application, use
  

For tomorrow

• What is a Knowledge Representation?
• Focus on the 5 roles
• Think about the logical vs. the psychological view of reasoning
• This class focuses on the logical
• What's the difference between sanctioning and recommending?
• What is the role of human beings?
• How do roles 4 and 5 relate?
• Analogize to a program
• How do roles 2 and 4 relate?
• How do roles 2 and 5 relate?