Task 3, Knowledge-based systems /Expert systems, Knowledge Representations
What are the four key features of an expert system, list and explain them?
* Expertise, which was explained above implies that the system must possess expert knowledge to make decisions as an expert of the field would. *Symbolic reasoning implies that knowledge must be represented symbolically, and decision be made using symbolic reasoning rather than mathematical calculation. (Backward and forward chaining). *Deep Knowledge implies that the system needs a certain level of knowledge. The knowledge needs to be at the level of an expert of the field. *Self-knowledge is something that the system needs to examine its own reasoning and motivate how it came to a certain conclusion. The system must have the ability to learn from its mistakes and success.
What is the difference between knowledge based systems and expert systems?
A Knowledge Based system is essentially an expert system but they use two different domains of data. A knowledge system generally works with information which currently is available, which may or may not be valid. On the other hand, an expert system works with reliable information which is known to be correct.
Explain some differences between knowledge based systems and a traditional computer based information systems.
A knowledge based system store information in the of facts and rules and uses those to reason and make conclusions while a traditional computer based information system simply stores information and can't produce anything except what it explicitly has been given.
What is a cognitive map?
A mental representation/map of one's environment, which can be represented as graphs or fuzzy cognitive maps in expert systems.
What is an expert system shell?
An expert system shell is an expert system with the knowledge removed. (Negnevitsky) Hence, everything that the user needs to do is to add certain knowledge to it, in the form of the rules, and also add specific data which is required to solve a problem. Hence, many expert system nowadays can be made by just one person filling in the data on the computer into the shell, that is, a team of 5 people, as mentioned in the general description above, is not required.
What is an ontology?
An ontology describes the relationship between different objects.
Explain backward chaining.
Backward chaining is an inference technique that starts with the goal. First it looks at the rules to see if it can find a rule that will result in the given goal, if not it tries to find a rule that requires the goal as a prerequisite and then works its way backwards to find to set of rules that will result in the goal.
Explain backwards chaining and give an example
Backwards chaining starts with facts being entered into a system and the system then using the facts to try and prove a hypothesis. If there are missing facts the system could use forward chaining which is the opposite of backward where you start with a hypothesis and work through the facts for it to help ask relevant questions for the user. Example: You know strawberries are red and small, and you know tomatoes are red and large. You then know the user has a red object so you ask if the object is small and then can come to a conclusion.
Give three examples of expert systems:
Chess solving system, taxation systems and MYCIN (medical diagnosis system).
Explain the difference between Cognitive maps and Inference networks.
Cognitive maps shows the relations between objects and concepts. Shows objects in a hierarchical graph where child objects have a link to their parent. Nodes have attributes/properties. Similar to Frames. Inference networks shows the rules of a system, children are premises and links are relations (AND/OR). So, both can show the objects and their properties, but they are presented in different ways.
Expert system
Designed to solve problems within a specific area and perform tasks that could be done by a human expert.
Can an Expert System replace a human expert?
Expert systems, when limited in their area of expertise, can perform exceptionally well. But they are only as good as the data loaded to it, e.g. IBM's famous Watson. Implicit knowledge, like tacit knowledge, cannot be applied by a computer. This leads to several problems, when the expert system reaches its predefined databased boundaries. E.g. an expert system used in an surgery-robot, how will the robot behave if it encounters a new situation which is not covered by the predefined data?
Explain forward chaining.
Forward chaining is an inference technique that starts with the data. The rules are first applied to the available data and the result is used to see if the goal was produced from this data.
Explain what a frame means, and how the frames are related in frame based system
Frame is a certain data structure which has specific information about a concept or object. Furthermore, every frame has its name and set of attributes. Furthermore, there are 3 types of relations in frame-based system: generalization, aggregation and association. Generalization represents a-kind-of relation between superclass and subclass. Aggregation represents a-part-of relationship where multiple components are associated in a superclass representing the whole. Association represents semantic relation between different classes which are, otherwise, independent.
What components form the knowledge base?
Heuristics, rules, objects, attributes, hypothesis, relationships, definitions, events, processes, facts.
Why is knowledge representation used in these systems?
In artificial intelligence "expert systems" are systems which can simulate the decision-making capability of an expert. It should be able to give advice within its specialized field. (lecture 3) Expert systems are a form of a knowledge-based system. There are two subsystems in an expert system. First, we have the knowledge base. In the knowledge base store all the rules and facts. Second, we have the interference engine. In the interference engine we apply the rules to the facts and deduce new information. Each new deduced piece of information can give the interference engine new rules and in return more information. (Wikipedia). Knowledge based systems is a very broad term. It refers to many systems who generates and uses knowledge. (Lecture 3) Knowledge representation is about how you are going to save knowledge. If the machine is going to use knowledge, we will need a way to store it. Depending on what kind of knowledge we wish to store, different structures may be suited. (lecture 3)
What are the limitations of expert systems?7
It can be difficult to extract all information needed from human experts, there may be facts that are known but the methods to derive the facts may not the as easy to represent in a knowledge based system.
What is Knowledge Representation?
It is the way the knowledge is stored in the knowledge base. For example, in a rule-based knowledge base where A and B are true, knowledge representation can be: IF A AND B THEN C. The knowledge of C is represented through the rule including A and B.
What are the limits of knowledge based systems and expert systems?
KBS - takes time to build, no common-sense, no intution, no thinking outside the box Expert systems - can't exchange knowledge base, only few thousand of rules, no commons sense.
What is O-A-V and how is it related to MYCIN?
O-A-V stands for Object - Attribute - Value and was used in the MYCIN system in this type of manner: (Ball, Color, Red) (Bob, Age, 22) But maybe with slightly more advanced medical terms. Either way, this way MYCIN could store only data that was relevant, and could leave knowledge that was missing empty. Instead of filling a large matrix or the likes with info.
Explain the relation between Expert systems and Knowledge-based systems.
One could say that knowledge-based systems are the framework, or the structure to tackle the problem, while expert systems refers to the specific use of the system. Almost all expert systems are knowledge-based systems, but not all knowledge-based systems are Expert systems.
What are typical problems at the knowledge elicitation and how can they be approached?
One problem could be that the expert gives controversial answers. In this case, asking other experts in this domain will be helpful. If different experts give different solving strategies for the same problem, confronting each of the experts with the contradicting strategies may result in aligned answers.
One knowledge representation, perhaps the most common, is production rules. Three problems that can arise are subsumed rules, circular rules and conflicting rules. Describe them and how you can solve them.
One rule can be subsumed by another if it is included in it. Handle it with meta-rules, so that the subsumed rules condition is a premise in the rule which subsumed it. Ex: R1: A && B => X R2: A && B && C => Y With meta-rule: MR2: R1 && C => Y Circular rules are (meta)rules which are independent of each other forming a cycle. They are avoided by not forming cycles, use some graph algorithm if too complex. Ex: R1: A && B => X R2: R1 && R3 => Y R3: R2 && C => Z Conflicting rules are rules which create a conflict and thus gives conflicting conclusions. Avoid conflicting rules by being consistent. Ex: R1: A && B => X R2: A && B => !X
What are production rules in Expert Systems?
Production rules are ways the system can determine a possible solution to a certain input data. If we take the earler example of MYCIN, that was a system that accepted a bunch of data and then compared it to certain rules to determine a diagnosis. Some examples for simpler diseases could be If HEADACHE then COLD, BRAIN TUMOR, POOR EYESIGHT If HAS_GLASSES then COLD, BRAIN TUMOR This way, the system can remove the diagnosees that are no longer possible.
Name some advantage with Knowledge-based systems.
Since the system is built upon explicitly represented knowledge, it is capable of explain its reasoning with full precision, unlike humans. By presenting the rules with which it drew its conclusion it can present combinations of reasoning which might have been previously unknown and unthought of.
Knowledge-based system
Stores rules and related data in a knowledge base. The system use the rules and the data to draw conclusions.[
Explain structural and meta knowledge
Structural knowledge shows how and in which way things work, and what is the relation between those things. On the other hand, meta-knowledge is considered to be knowledge about knowledge, and the concept of meta-knowledge is specifically useful when we need to decide which knowledge should be used in order to solve the problem in the best way.
Explain the concept of Tacit knowledge
Tacit knowledge is considered to be a specific kind of knowledge that is considered to be difficult to express or pass on to other people. Typical example of tacit knowledge is the ability to cycle bikes. It is not fully known how we keep balance, but we do know that. Furthermore, the opposite of tacit knowledge is called explicit knowledge.
Explain the KR scheme Frame and some of its advantages.
The frame scheme is an object-oriented data structure. Each frame consists of a set of slot - value pairs where the slot describes what kind of information is stored in the value. Frames can also be connected in a way where one frame is a subset of another frame and it then inherit properties from its parent frame that then can be changed if needed. One of the advantages of frames is their flexibility as it can easily deal with exceptions by adding new slot - value pairs to a given frame.
What are some known disadvantages with the Expert systems?
The main disadvantage is the knowledge acquisition needed within the specific domain to be able to act as a human expert. Experts who are qualified enough within the domain are almost certainly few and their time, per definition, highly valued elsewhere.
What is the main purpose of Expert Systems?
The main purpose of Expert Systems is to replicate knowledge and skills of human experts in a particular area, and then to use this knowledge to solve similar problems without human experts participation. In short, expert systems should be able to help a novice to make the same decision and draw the same conclusion that an actual expert would make or possibly exceed.
Are expert systems also knowledge-based systems?
The term knowledge-based system describes only the kind of a architecture of a system. Expert systems are based on the domain knowledge of experts to replace or aid experts in the future process and are therefore considered as specified knowledge-based systems.
Name the two subsystems in an expert system and what they do
The two subsystems are the knowledge base and the interference engine. The knowledge base stores the rules and facts. The interference engine uses the knowledge base to deduce new information from those rules and facts.
What are the limitations of expert systems?
● They can only handle specific domain compared to a human expert which can handle more domains. ● They don't have common sense or emotions to be aware of the situation. ● Expert systems can't learn for themselves and needs to manually receive new data.