TOPIC 6 - EXPERT SYSTEMS (EIS) Flashcards
Define what an ES is.
ES is essentially a computerized service that provides the knowledge that users typically expect from a human expert.
In other words, expert knowledge is integrated into the system in order to provide automated responses to queries. This could be in the form of high-level legal advice, information from a chiropractic therapist and so forth. Information in the system is regularly updated, thus, such systems seldom become outdated.
The knowledge base that an ES relies on to provide advice is built through a reasoning process which is handled by an inference engine – the systems thinking component. The inference engine is essentially an algorithm programmed to use a deduction process or backward chaining technique to analyse problems or queries presented to the system. A simplified explanation of this technique is as follows:
An analysis begins with the hypothesis of the „most likely scenario for the relevant requested information and based on data available in the system. When the hypothesis cannot be supported, the system will move to the „next most likely scenario and the process continues until it succeeds in finding the solution.
This process is very similar to the process of medical diagnosis. In fact, some of the earliest applications of ES came out of the medical field. In the medical context, a physician will begin with what he believes to be the most likely cause and proceed to carry out tests to confirm it. If the result is not conclusive, it would be necessary to carry out further tests. On the other hand, if there are contradictory findings to the initial diagnosis, other diagnoses will be required. This will go on until a confirmed finding is achieved.
Explain three functions that characterise an ES.
The ES is generally embedded with the following functions;
(a) Interactivity
An ES should be an interactive and computer-based tool whereby the user can query and be queried or prompted by the system for more information when necessary in order to solve a problem.
(b) Heuristics
An ES should use a combination of facts and guesswork to solve a problem. The latter involves a process called heuristics whereby the system makes an approximation in situations where the information is either unavailable or takes too long to find. This enables the system to provide a solution even though it is without complete or accurate data but within a reasonable time frame.
(c) Simulation
Response from an ES should be dependable or trustworthy enough for the user to perceive it as a computer simulation or computerised representation of an expert.
State four benefits of using ES.
In the real world, feedback is very important as it can help a good expert grow by learning from it. When the feedback is incorporated into the knowledge base, the expert system will become „smarter‰ too. How feedback is handled by the different components of ES is also different and can be classified according to the level of dynamism in each componentÊs response to the feedback.
(a) For Working Memory: Most Dynamic
The working memoryÊs contents not only change with each different problem situation that the ES handles, but also with each step that is taken to resolve the problem. The data structure in the working memory is kept most current and is, therefore, the most dynamic component of the ES.
(b) For Knowledge Base: Moderately Dynamic
When a new piece of information arises which invokes a change in the problem solution procedure, the knowledge base would be required to change as well. It must be noted that before the changes can be implemented, they are also carefully evaluated by the system. This means that the changes must not be entirely based on only one specific experience during the consultation, since a rule considered irrelevant in that particular situation might just be crucial in others.
(c) For Inference Engine: Least Dynamic
Changes will only be made if it is absolutely necessary for debugging or when the inferential process needs to be enhanced. This is because the inference engine has very strict control and coding structures. Meanwhile, in commercial inference engines, changes are made at the developerÊs discretion. Changes are usually minimal because too frequent updates would incur cost to the clients and can be very disruptive.
Describe the functions of:
(a) Knowledge base; and
(b) Inference engine.
(A) This is where the knowledge of the system is stored. It is where the facts and rules of the expert domain are kept and retrieved from in order to solve problems. In the database of rules, the context of the problem domain will be provided. This is done through a set of useful facts that could be used in the IF-THEN rules model.
(B) The inference engine is the thinking part of the system. It is tasked to control how the IF-THEN rules are used. During a problem-solving session, acquisition of more information from the user would be allowed. He would usually be prompted to provide more input through an interface that uses natural language. In this part of the expert system, a reasonable solution could be derived through the set of answers provided by the user. Here, perhaps the simplest inference engine could be in the form of a forward-decision tree.
