1 / 42

Knowledge-Based Expert Systems

Knowledge-Based Expert Systems.

langston
Download Presentation

Knowledge-Based Expert Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Knowledge-Based Expert Systems A computer program which, with its associated data, embodies organised knowledge concerning some specific area of human activity. Such a system is expected to perform competently, skilfully and in a cost-effective manner; it may be thought of as a computer program which mimics the performance of a human expert.

  2. Knowledge-Based Expert Systems The Nature of Expertise One view of human expertise is that some people have spent so much time solving problems in one particular domain that they ‘know all there is to know’ (nearly) and are able to see any problem as an instance of a class of problems with which they have been confronted before. Once the expert has successfully classified or recognised a new problem as an instance of a previously experienced problem type, all the expert has to do is apply whatever solution proved successful in dealing with that type of problem in the past.

  3. Knowledge-Based Expert Systems The Nature of Expertise This model of human expertise that relies (a) on domain-specific knowledge and (b) experience-based recognition of solutions of problems and has served as the basis of numerous expert systems. The idea is that cognition is a ‘recognition’-based phenomenon. That is, when a medical doctor, say, examines a patient or hears what the patient has to say about his or her problem, the configuration of symptoms or signs suggests a particular illness with which the doctor is already completely familiar.

  4. Knowledge-Based Expert Systems The Nature of Expertise The ‘recognition-based’ phenomenon can be viewed as setting up a key goal in the problem solving process and then attempting to find out data that satisfies the goal. The goal is then broken up into sub-goals and data sought to prove each of the sub-goals. (The sub-goals can be broken into further sub-goals and so on and on). While proving a sub-goal, the experts suspend the goal and try and satisfy individual goals. Once all sub-goals are satisfied then the key goal is deemed to be satisfied and the problem solved!

  5. Knowledge-Based Expert Systems Pneumonia Expert System Consider a small chunk of a medical experts knowledge about diagnosing pneumonia and fever rulepneumonia if ‘the patient has chest pain’ & ‘the patient has a fever’ & ‘the patient produces purulent sputum’ then ‘the patient has pneumonia’ rulefever if ‘the patient has a temperature above 100’ then ‘the patient has a fever’. The nurse has just come in a with a patient with the following symptoms: ‘the patient has a chest pain’ & ‘the patient has a temperature above 100’& ‘the patient produces purulent sputum’ And the expert has to deduce whether or not the patient has pneumonia?

  6. Knowledge-Based Expert Systems Pneumonia Expert System The nurse has just come in a with a patient with the following symptoms: ‘the patient has a chest pain’ & ‘the patient has a temperature above 100’& ‘the patient produces purulent sputum’ And the expert has to deduce whether or not the patient has pneumonia? In order to prove whether or not the patient has pneumonia, the expert has to prove: prove ‘the patient has pneumonia’  GOAL prove ‘the patient has a chest pain’  SUB-GOAL prove ‘the patient has a fever’ SUB-GOAL prove ‘the patient produces purulent sputum’ SUB-GOAL

  7. Knowledge-Based Expert Systems • Intelligent beings perceive, reason and act.  • Intelligent beings are creative, learn from their mistakes.  • Intelligent beings can learn from their environment.  • Intelligent beings can learn with the help of tutors.  • Intelligent beings can work on their own/form groups.  • Intelligent beings have a value system, an exchange system.

  8. Knowledge-Based Expert Systems MYCIN, microbial infection therapy system comprised over 400 rules of thumb such as below which help in microbial infection therapy, for example in the diagnosis and therapy of meningitis. IF: 1) The stain of the organism is grampos, and 2) The morphology of the organism is coccus, and 3) The growth conformation of the organism is chains THEN: There is suggestive evidence (0.7) that the identity of the organism is streptococcus. Each of the antecedent clauses and the consequent clause is essentially a complex relationship between the terms of this subject: the stain, morphology and growth, conformation of an organism and its identity and the attributes of these = grampos, coccus, chains, streptococcus. The rules show the interrelationship between various domain objects ('the stain' etc.) and how such a relationship could be used to infer new facts. The objects and their attributes, (for example, "morphology" and its attribute "coccus") viewed here as a data structure, are each viewed by the terminologist as terms representing concepts.

  9. Knowledge-Based Expert Systems The expert system R1/XCON was used by DEC(now a part of COMPAQ) for configuring computer systems. This system was in routine commercial use and contains over 10,000 rules of the type shown below IF: The most current active context is assigning a power supply & an sbi module of any type has been put in a cabinet & the position it occupies in the cabinet (its nexus) is known & there is space available in the cabinet for a power supply for that nexus & there is an available power supply THEN: put the power supply in the cabinet in the available space. The above rule is much more complex than the rule shown for MYCIN in that although we are talking about putting the power supply in the cabinet in the available space, each of the domain objects and its attributes is a complex entity in itself: for example an SBI moduleof any type, the power supply and the available space in the cabinet.

  10. Knowledge-Based Expert Systems

  11. U S E R I N T E R F A C E A L G O R I T H M (S) D A T A B A S E Data Facts C O M P U T E R S Y S T E M COMPUTER SYSTEM Knowledge-Based Expert Systems A conventional computing system

  12. Data Facts Knowledge-Based Expert Systems A Knowledge-Based system U S E R I N T E R F A C E H E U R I S T I C S & A L G O R I T H M (S) K N O W L E D G E B A S E Rules Meta Rules C O M P U T E R S Y S T E M COMPUTER SYSTEM

  13. Knowledge Base An organised and structured repository of ‘human knowledge’ in a given specialist domain. This repository can be updated or deleted in parts. Usually the knowledge is encoded as conditional ‘if <antecedent> then <consequent>‘ statements (‘rules’ and ‘problem-solving tasks’) together with the so-called ‘domain objects’. The latter are descriptions of facts, including concrete and abstract facts, relating to the specialist domain.

  14. Knowledge Base Knowledge representation is about making things explicit, is about resolving ambiguities; Knowledge representation, in the context of artificial intelligence, is about describing a class of things to a computer system. This description should not be ambiguous either lexically or structurally This description should explicate shared knowledge

  15. Reasoning Strategies • Reasoning may be characterised as an attempt to combine elements of old information to form new information. • Reasoning strategies refer to the rather long sequences of individual small inferences organised so as to address a main goal or problem. • Reasoning strategies involve the representation of information and knowledge, the use of inference rules for manipulating that knowledge & information, and control mechanisms for making the variety of choices necessary in the search for solutions.

  16. Problem Solving • Problem-solving is sometimes defined as a process that involves finding or constructing a solution to a problem. • Human problem solving can be modelled as the exploration of different paths to a solution and involves ‘information processing’ which appears unique to human beings. • Cognitive psychologists typically divide problems into two classes: ‘well defined’ or closed world problems including solutions of games or puzzles and ‘ill defined’ or open world problems.

  17. Heuristics or Rules of thumb Prevent the hi-jacking of airliners Prevent hi-jackers from boarding the airliners Heuristic technique Algorithmic route • Put passengers and luggage through a metal detector & • Search only those who set off the detector • Search those passengers that match a predetermined hi-jacker profile • (inc. passengers, flight crews & mechanics) • Strip search every person with access to the airlines & • Search all luggage

  18. Knowledge Based Systems & Conventional Systems Conventional Programs KBES Representation data knowledge Reasoning algorithmic & repetitive heuristic & inferential Retrieves large DB large KB Knowledge encrypted represented

  19. Knowledge Based Systems & Applications • There are recognised experts • The experts are provably better than novices • The tasks takes an expert a few minutes to a few hours (if it takes days - FORGET IT) • The task is primarily cognitive • The skill is routinely taught to novices • The task requires no common sense • The task domain is important: economically, financially or socially

  20. Knowledge Engineering The accumulation, codification and application of knowledge through the use of computer systems, specifically knowledge-based systems.

  21. Knowledge Engineering

  22. Rule-based Systems • A rule-based system helps us to codify the problem-solving knowledge of the human expert. • It appears that experts typically express their knowledge as a set of situation-action rules. • RBS research should address the need to capture, represent, store, distribute, reason about and apply human knowledge electronically. Hayes Roth, F. (1992). ‘Rule-Based Systems’ p.1426

  23. Information Exchange & Natural Language Natural Language. A person’s native tongue. Natural Language interface. A system for communicating with a computer by using a natural language. Natural Language Processing. Processing of natural language (e.g., English) by a computer to facilitate communication with the computer or for other purposes, such as language translation.

  24. Information Exchange and Visual I/O • Recognising and reasoning about the visual environment something that people do extraordinarily well; • In these abilities an average three year old makes the most sophisticated computer vision system look embarrassingly inept

  25. Information Exchange and Visual I/O

  26. Representation: Production Systems • Production Systems are a modular knowledge representation scheme and are based on the notion of condition-actionpairs, called production rules or just productions: "If this condition occurs, then do this action". • The utility of the production system formalism comes from the fact that the conditions in which each rule is applicable are made explicit and, in theory at least, the interactions between rules are minimised in the sense that the rules do not 'call' each other.

  27. Representation: Production Systems Consider a knowledge-base containing the following rules: Rule#1:IFA&B&CTHEN D Rule#2: IFD&FTHEN G Rule#3: IFA&JTHEN G Rule#4: IFBTHEN C Rule#5: IFFTHEN B Rule#6: IFLTHEN J Rule#7: IFGTHEN H. The knowledge-base also contains the following facts Fact#1: A. (‘A’ is known to be true) Fact#2 : F. (‘F’ is known to be true)

  28. Representation: Production Systems

  29. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: FIRST CYCLE

  30. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: SECOND CYCLE

  31. Representation: Production Systems Consider a knowledge-base containing the following rules: Rule#1:IFA&B&CTHEN D Rule#2: IFD&FTHEN G Rule#3: IFA&JTHEN G Rule#4: IFBTHEN C Rule#5: IFFTHEN B Rule#6: IFLTHEN J Rule#7: IFGTHEN H. The knowledge-base also contains the following facts Fact#1: A. (‘A’ is known to be true) Fact#2 : F. (‘F’ is known to be true)

  32. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: THIRD CYCLE

  33. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: FOURTH CYCLE

  34. Representation: Production Systems Consider a knowledge-base containing the following rules: Rule#1:IFA&B&CTHEN D Rule#2: IFD&FTHEN G Rule#3: IFA&JTHEN G Rule#4: IFBTHEN C Rule#5: IFFTHEN B Rule#6: IFLTHEN J Rule#7: IFGTHEN H. The knowledge-base also contains the following facts Fact#1: A. (‘A’ is known to be true) Fact#2 : F. (‘F’ is known to be true)

  35. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: FIFTH CYCLE

  36. Representation: Production Systems An Example Problem: To prove that H is true?  Cycles of Production: SIXTH CYCLE

  37. Representation: Production Systems Conflict Resolution Rule Ordering Arrange rules in list with most important rules higher up the list. Fire rules according to their position in list (highest first). Context Limiting Place rules in groups, and only have one group ‘active’ at any one time. Specificity The specificity principle states that if a number of rules are applicable to a given situation then the rule with the greatest number of condition premises (I.e. the most specific) should be selected to fire.

  38. Representation: Production Systems Conflict Resolution Refractoriness Refractoriness is another conflict resolution strategy which states that if a rule has been applied on a previous cycle, then it should not be applied again to the same set of facts in data memories. This kind of strategy prevents a system from getting itself entwined in a loop. Recency The recency principle is a conflict resolution strategy which states that if more than one rule applies to a given situation, then choose the rule that applies to the most recently entered items in data memory.

  39. Deduction KBS Problem: Elicit rules from the following description Johnny is an amateur zookeeper and also keeps notes in his diary on the various animals and birds he comes in contact with. His diary contains the following entries: 'Jan 1, 1992: Mary told me that meat-eating animals with pointed teeth, forward pointing eyes and claws are called carnivores. Like other mammals they give milk.' 'Feb. 15, 1992: I saw two tawny coloured carnivores today, but one had dark spots and the other black stripes. The dark spotted carnivore was called cheetah and the black striped was a tiger.' 'March 20, 1992: We had two new animals delivered to the zoo, a zebra and a giraffe. These hoofed mammals are called ungulates. The zebra has white skin with black stripes on it. The giraffe had long legs and neck, and has the same tawny colour as the tiger but with black spots'.

  40. Deduction KBS

  41. Deduction KBS Johnny has been asked to identify a hoofed, long-necked hairy animal, Freddy, which has a tawny colour and dark spots. Use the forward chaining search strategy to draw the inference network which Johnny should use to identify the unknown animal. Fired 1st Gives milk Is a mammal Fired 2nd Has hoofs Is an ungulate Has long legs Fired 3rd Has a long neck Is a giraffe Has a tawny colour Has dark spots

  42. Deduction KBS

More Related