Knowledge Representation Other Examples Volker Sorge Intro to AI: - - PowerPoint PPT Presentation

knowledge representation
SMART_READER_LITE
LIVE PREVIEW

Knowledge Representation Other Examples Volker Sorge Intro to AI: - - PowerPoint PPT Presentation

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Knowledge Representation Other Examples Volker Sorge Intro to AI: Knowledge Representation Lecture 5 Volker Sorge Background Object Oriented


slide-1
SLIDE 1

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Knowledge Representation

Volker Sorge

slide-2
SLIDE 2

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Knowledge Representation

◮ The problem of how to represent knowledge and

reasoning with it or about it.

◮ Obvious questions are:

  • 1. What is knowledge?
  • 2. What is a representation?
  • 3. How do we reason?

We are mainly interested in the first two questions here.

slide-3
SLIDE 3

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Implicit vs. Explicit

◮ Knowledge can be learned and implicitly represented. ◮ Examples: Neural nets, evolutionary algorithms. ◮ Knowledge can be explicitly represented. ◮ Allows for human encoding, extension and inspection. ◮ We can learn something about the knowledge, explain

behaviour etc.

slide-4
SLIDE 4

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Types of Explicit Knowledge Representation

◮ Logic representation

◮ Knowledge is represented as logical sentences. ◮ Each sentence is self-contained ◮ But representation is flat

◮ Object Oriented representation

◮ Hiearchically structured and organised ◮ Sentences are grouped ◮ Translatable to logic

slide-5
SLIDE 5

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Semantic Networks

◮ Hierarchies of Classes similar to object oriented

programming.

◮ Classes are sets of entities. ◮ Classes have properties that can be inherited. ◮ Both classes and entities can have relations. ◮ Multiple inheritance is always conjunctive.

Expresses simple sentences of Object – Predicate – Subject:

◮ A bird is a vertebrae. ◮ Paul is a brother and a father.

slide-6
SLIDE 6

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Simple Semantic Networks

◮ Semantic Networks allow for simple subclass relations

(IsA links) and properties (HasA links).

Source: Wikipedia

slide-7
SLIDE 7

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Complex Relations

◮ More complex models can incorporate more relations.

Source: http://www.jfsowa.com/pubs/semnet.htm

slide-8
SLIDE 8

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Problems

◮ Inheritance can lead to incorrect knowledge. ◮ Example:

◮ Bird has the property flight. ◮ Penguin inherits from bird.

◮ Multiple inheritance can aggravate this problem and

lead to contradiction.

◮ Solution: Override inheritance, where sub-class values

  • verride super-class values.
slide-9
SLIDE 9

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

RDF

◮ W3C resource description framework for semantic web

applications

◮ Basic idea is to enable a semantic description of

information implemented as web resources

◮ Automatic identification of different web services ◮ Sharing of data across different resources ◮ Also enables easier retrieval of similar facts

slide-10
SLIDE 10

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

RDF: Basics

◮ Resources are given by URIs ◮ Described in Object – Predicate – Subject relations ◮ This effectively defines a graph with directed edges for

relations

◮ Example:

:Birmingham a :City :Birmingham :inCounty :WestMidlands :Birmingham :inCountry :UK :UK :inContinent :Europe All subjects and objects could be described with Maps URLs.

slide-11
SLIDE 11

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

RDF Example: Google recipe search

◮ Webmasters/Authors are encouraged to use semantic

markup for recipes

◮ Microdata or RDF format for ingredients, cooking time,

etc.

◮ This enables Google search for recipes with respect to

ingredients only.

◮ https://developers.google.com/

structured-data/rich-snippets/recipes

<span rel="v:ingredient"> <span typeof="v:Ingredient"> <span property="v:name">White sugar</span>: <span property="v:amount">3/4 cup</span> </span> </span>

slide-12
SLIDE 12

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Example: Wolfram Alpha

◮ http://www.wolframalpha.com ◮ (Mainly) scientific knowledge base. ◮ Scientific backend of Siri. ◮ http://www.wolframalpha.com/examples

slide-13
SLIDE 13

Intro to AI: Lecture 5 Volker Sorge Background Object Oriented Representations Example: RDF Other Examples

Example: Google Knowledge Graph

◮ http://www.google.co.uk/insidesearch/

features/search/knowledge.html

◮ (Now only available on the wayback machine!) ◮ Scientific and general knowledge base. ◮ Mined from Wikipedia and search.