Designing Expert Systems
Designing Expert Systems CS 681
Popular in Course
verified elite notetaker
Popular in ComputerScienence
This 52 page Class Notes was uploaded by Leland Swift on Monday September 28, 2015. The Class Notes belongs to CS 681 at George Mason University taught by Gheorghe Tecuci in Fall. Since its upload, it has received 25 views. For similar materials see /class/215119/cs-681-george-mason-university in ComputerScienence at George Mason University.
Reviews for Designing Expert Systems
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/28/15
UNIVERS39 C3681Fall2008 Gheorghe Tecuci tecucigmuedu httplacgmuedul Learning Agents Center and Computer Science Department George Mason University 2008 Leaming Agents Center Questions and Exercises Briefly define and compare data information and knowledge with the help of an example of each of these concepts Which are the different uses of an expert system Why it is hard to build an expert system Describe very briefly 6 basic concepts elicitation methods Which are the main strengths and weaknesses of these methods Describe briefly the cardsort elicitation method and specify its main strengths and weaknesses In its standard form the cardsort method elicits a strict hierarchy of concepts How could one modify this method to build a tangled hierarchy 2008 Leaming Agents Center Questions and Exercises Analyze the following dialog and develop an ontology concepts feature definitions instances and facts suggested by it KE Suppose you were told that a spill had been detected in White Oak Creek one mile before it enters White Oak Lake What would you do to contain the spill SME That depends on a number of factors I would need to find the source in order to prevent the possibility of further contamination probably by checking drains and manholes for signs of the spill material And it helps to know what the spilled material is KE How can you tell what it is SME Sometimes you can tell what the substance is by its smell Sometimes you can tell by its color but that39s not always reliable since dyes are used a lot nowadays Oil however floats on the surface and forms a silvery film while acids dissolves completely in the water Once you discover the type of material spilled you can eliminate any building that either don39t store the material at all or don39t store enough of it to account for the spik 2008 Leaming Agents Center Questions and Exercises H II Briefly define verification validation and certification Describe very briefly the main stages of building an expert system Briefly describe three limiting factors in building expert systems What is a expert system shell What is a learning system shell Specify the sequence of phases in the development of an expert system Specify the sequence of phases in the development of an expert system with the Disciple shell Briefly compare the classical approach to the development of an expert system with the approach based on using a learning agent shell 2008 Leaming Agents Center Questions and Exercises What is an object ontology Which are the different uses of an object ontology in a knowledge based learning agent Consider a knowledgebased agent Compare the relative generality of its object ontology and its rules What is an instance What is a concept Give an intuitive definition of generalization What does it mean for conceptA to be more general than concept B 2008 Leaming Agents Center Questions and Exercises What are the possible relationships between two conceptsA and B from a generalization point of view Provide examples of concepts A and B in each of these three situations How could one prove thatA is more general than B Is this always a practical procedure How could one prove thatA is not more general than B Is this a practical procedure Briefly describe the guidelines related to siblings for designing an ontology Design an ontology fragment concepts feature definitions instances and facts to represent the following information John Doe has written the Windows of opportunities book between 2005 and 2007 2008 Leaming Agents Center Exercise Consider the following feature hierarchy Is there any necessary relationship between BD1 and 1D1 F t 1 ewe m BR1 and 1R1 A2D1 and 1D1 9 mm m AD1 and BD1 1D1 and 1R1 DOMAIN eatu re A1 CFeatu re A2 2008 Leaming Agents Center Questions and Exercises Using the ontology design pattern for concepts and subconcepts represent the following information in a graphical form Support is a PhD advisor criterion It has as basic subcriteria ie criteria that do not have subcriteria the following advisor funding advisee AY support advisee summer support advisee funding assistance and advisee conference support Consider the evaluation criterion concept and one of its instances called advisee AY support from John Doe during the AY0809 Represent in an ontology the various features of this instance including the fact that this support is very high Briefly discuss with the help of an example why maintaining the consistency of the object ontology is a complex knowledge engineering activity 2008 Leaming Agents Center Questions and Exercises Briefly explain how the following questions are asked assuming the knowledge from the following ontology What is the retirement age of John Smith What is the retirement age of Tom High What is the retirement age of Jane Austin faculty member PhD advisor retirement 66 instructor assistant associate full retirement age 70 professor professor professor inst nceof instarc of in tance Oohn SmitlD John Doe Jane Austin Cl39om HigrDremem m 66 2008 Leaming Agents Center Questions and Exercises Considerthe following modification of an ontology Explain why if we delete the subconcept of relationship between B to A C can no longer have the feature f COM 6 COM 69 Initial State Modi ed State 2008 Leaming Agents Center 10 Questions and Exercises Consider the following representation featu re1 stance1 QnstanceZD stanceZD Which are all the facts that can be inferred from it 2008 Leaming Agents Center 11 Questions and Exercises Develop an object ontology that represents the following information Puss is a calico Herb is a tuna Charlie is a tuna All tunas are fishes All calicos are cats All cats like to eat all kinds of fish Cats and fishes are animals Hint You should define object concepts object features and instances 2008 Leaming Agents Center Questions and Exercises Develop an object ontology that represents the following information The government of US is a representative democracy The government of Britain is a parliamentary democracy George W Bush is the head of the government of US George W Bush has a critical role in setting objectives for US You should define object concepts object features and instances keeping in mind that you will need to extend this ontology with new knowledge in the future 2008 Leaming Agents Center Questions and Exercises object subcon eptof Type of government Headof government country Schoncep39 Instan eof Parliamentary Representative US democracy democracy Instan eof HascriticaIrole Instanceof Insta ceof in39setting39 objectivesfor Isheadof Government of Britain Government of US George W Bush feature domain object ranvge ObjeCt Isheadof Hascriticaroeinsettingobjectivesfor domain head of government domain head of government range type of government range country 2008 Leaming Agents Center 14 Questions and Exercises Insert the additional knowledge that platypus lays eggs into the following object ontology 1 2008 Leaming Agents Center Questions and Exercises What is a positive example of a concept What is a negative example of a concept What is a generalization rule What is a specialization rule What is a reformulation rule Name all the generalization rules you know Briefly describe and illustrate with an example the turning constants into variables generalization rule Define and illustrate the dropping conditions generalization rule Indicate three different generalizations of the sentence History books in the Fenwick Library and demonstrate that each of them is more general than the given sentence 2008 Leaming Agents Center 16 Questions and Exercises Define and illustrate the following generalization rule climbing generalization hierarchies Solution An expression can be generalized by replacing a concept with a more general concept from a generalization hierarchy For instance the statement GMU PhD students are hard working can be generalized to GMU graduate students are hard working by replacing the concept GMU PhD students with the more general concept GMU graduate students 2008 Leaming Agents Center 17 Questions and Exercises Define and illustrate the following a generalization of two concepts b minimally general generalization of two concepts 0 maximally general specialization of two concepts What is a negative exception What is a positive exception Draw a picture representing a plausible version space as well as a positive example a negative example a positive exception and a negative exception Then briefly define each of these elements 2008 Leaming Agents Center 18 Questions and Exercises Define and illustrate the least general generalization of two concepts Does it always exist Answer G is the least general generalization of two conceptsA and B if and only if G is a generalization ofA and B and G is less general than any other generalization ofA and B Two concepts may not have a least general generalization In such a case there are several minimal generalizations ofA and B but none of them is less general than all the others 2008 Leaming Agents Center Questions and Exercises Consider the cells consisting of two bodies each body having two attributes color which may be yellow or green and number of nuclei 1 or 2 The relative position of the bodies is not relevant because they can move inside the cell You should assume that any generalization of a cell is described as a single pair s t u v a Indicate all the possible generalizations of the following cell and the generalization relations between them 1 green 2 yellow b Determine the number of the distinct sets of instances and the number of concept descriptions for this problem c Given the following cell descriptions 1 green 1 green 1 yellow 2 green 1 green 2 green Determine the following minimal generalizations gE1 E2 gE2 E3 gE3 E1 gE1 E2 E3 2008 Leaming Agents Center 20 Questions and Exercises Which is the set of instances represented by the following concept O1 instance of PhD student is interested in 702 702 instance of PhD research area Indicate such an instance 2008 Learning Agents Center 21 Questions and Exercises What does the following concept represents O1 instance of course has as reading 702 702 instance of publication has as author 703 703 instance of professor Which is an instance 2008 Leaming Agents Center 22 Questions and Exercises What does the following concept represents O1 instance of PhD student is interested in 02 O2 instance of PhD research area Except When 02 instance of PhD research area requires programming 2008 Leaming Agents Center Questions and Exercises Indicate several generalizations of the following sentence Students who have majored in Computer Science at George Mason University between 2003 and 2004 Provide another example of a concept and indicate some of its generalizations Indicate several specializations of the following sentence Students who have majored in Computer Science at George Mason University between 2003 and 2004 Provide another example of a concept and indicate some of its specializations 2008 Leaming Agents Center 24 Questions and Exercises Give an example of a natural language sentence C that has some concept interpretation Formulate a sentence G which is a generalization of C and use the generalization rules to demonstrate that G is a generalization of C Formulate a sentence 8 which is a specialization of C and use the generalization rules to demonstrate that S is a specialization of C Formulate a sentence D which is neither a generalization of C nor a specialization of C 2008 Leaming Agents Center 25 Questions and Exercises Demonstrate that C is more general than C1 c1 01 is assistant professor 39 number of publications 10 is employed by George Mason University C 01 is professor number of publications N1 N1 is in 10 35 2008 Leaming Agents Center 26 Questions and Exercises Demonstrate that C is a generalization of C1 and CZ C1 01 is assistant professor numberof publications 10 is employed by George Mason University C201 is associate professor number of publications 35 C 01 N1 is professor numberof publications N1 is in 10 35 2008 Leaming Agents Center Questions and Exercises Consider the following examples E1 Determine their minimally general generalizations assuming the ontology from the 39201 instance of is interested in 02 instance of graduate research assistant 02 PhD research area 701 702 instance of is interested in instance of teaching assistant 02 PhD research area next slide Determine a common generalization which is not a minimally general generalization Is there a least general generalization Consider the following generalizations Determine their maximally general specializations assuming the ontology from the G1 39201 instance of is interested in 02 instance of university employee 02 research area 39201 instance of is interested in 02 instance of graduate student 02 research area next slide Determine a common specialization which is not a maximally general specialization 2008 Learning Agents Center Questions and Exercises person rect subconcept 0 university employee student Ctsu one f directs co eptof graduate Gtaff membeD culty membeD StUdent ndergraduate d39 su o c tof StUS e direct sub oncept of graduate teaching research assistant assistant PhD student MS student in n eof insta ce of Jane Austin ob Sharg Qoan Dea 2008 Leaming Agents Center 29 Questions and Exercises Which are some concepts included in this version space Universe of instances Concept to x be learned 4 Plausible Upper Bound Plausible Upper Bound O1 instance of faculty member student is interested in 02 O2 instance of research area Plausible Lower Bound O1 instance of graduate student is interested in 02 O2 instance of PhD research area associate professor 2008 Leaming Agents Center Questions and Exercises Consider the following partially learned concept and 10 instances Order the instances by the plausibility of being positive examples of this concept and justify the ordering 2008 Leaming Agents Center Questions and Exercises Define the statebased representation of a problem Define the reduction representation of a problem What are some of the complementary abilities of humans and computers What are some of the complementary abilities of humans and computer assistants in the context of webpage believability assessment What is mixedinitiative reasoning 2008 Leaming Agents Center 32 Questions and Exercises What is a problemreduction rule What is a solutionsynthesis rule 2008 Leaming Agents Center Questions and Exercises Consider the following ontology fragment publication subco cepz of subconceptof subco cepz of I insta ceof Masoncs430 has as reading Doe 2000 has as author university course of as as reading Explain how the following question will be answered Is there a course that has as reading a publication by a professor 2008 Leaming Agents Center 34 Questions and Exercises Consider the following ontology fragment where the unlabelled links are instanceof or subconcept of links 3 W WW WM MM cleaner loudsp eakercomD CsoftleaneD ardcleaneDLwaCsoftcomp one ardcomp one z X a eme ae 2008 Leaming Agents Center 35 Questions and Exercises Consider the following question Is there a cleaner X that removes dust Represent the question as a network fragment Find all possible answers to the question based on the information from the above ontology fragment In order to answer the question the agent would need to use several reasoning operations Which are these operations 2008 Leaming Agents Center 36 Questions and Exercises Consider the following ontology fragment where ISA means subconcept of quotquotmwmeOUDSPEKER PART OF AKEHPONE ISA ISA ISA OFIEMBR NE ISA CHASS COLORINSTAPOF EMBR E ASSEMLY PART OF ISA INSTANOF I CONTAINS PROVIDER j GLUES GLUES MADE OFINSTANOF ISA E INSTANOF PROVIDER MECHANIEHRSSlef I GLUES39 MADE OF OWICmL ISA MADEOF INSTAIOF CHASSJASSEM INSTAI OF INSTAIOE ICAOUTCHOUEONTMWHESDIVE LY 2008 Leaming Agents Center 37 Questions and Exercises Consider the question Is there a part of a loudspeakerthat is made of metal In the context of the object ontology from the previous slide a Which are all the answers to this question b Which are the reasoning operations that need to be performed in order to answer this question 0 Consider one of the answers that requires all these operations and show how the answer is found 2008 Leaming Agents Center 38 Questions and Exercises Consider also the following expressions in the context of the previous ontology fragment E1 X IS MEMBRANE E2 X IS MECHANICALCHASSIS MADEOF M MADEOF M M IS PAPER M IS METAL Z IS CONTACTADHESIVE Z IS MOWICOLL GLU ES M GLU ES M STATE fluid a Find the minimally general generalizations of E1 and E2 Solution MGG I MGG2 X Is LOUDSPEAKERCOMP X Is LOUDSPEAKERCOMP MADEOF M MADEOF M M Is MATERIAL M Is MATERIAL 72 Is lNFLAMMABLEOBJECT 72 Is ADHESIVE GLUES M GLUES M 2008 Leaming Agents Center 39 Questions and Exercises b Find two generalizations of E1 and E2 that are not minimally general generalizations G1 G2 X IS SOMETHING X IS SOMETHING MADEOF M MADEOF M M IS MATERIAL M IS SOMETHING Z IS INFLAMMABLEOBJECT Z IS SOMETHING GLUES M GLUES M c Consider one of the generalizations found at b and demonstrate why it is a generalization of E1 and E2 but it is not a minimally general generalization A MGG of E1 and E2 must not be more general than any other generalization of E1 and E2 G1 is not a MGG because is is more general than for example MGG1 from answer a This is true because XSOMETHNG is more general in the semantic network than X LOUDSPEAKERCOMPONENT d Is there a least general generalization of E1 and E2 No because the LGG is by definition the one and only MGG and we have found MGG1 and MGG2 2008 Leaming Agents Center 40 Questions and Exercises e Indicate a specialization of E1 S1 X IS MEMBRANE MADEOF M M IS PAPER Z IS CAOUTCHOUC GLUES 7M STATE fluid 2008 Leaming Agents Center Questions and Exercises Consider the following rule and ontology fragment Explain how the following problem is reduced Assess Bill Jones as a potential PhD advisor for Dan Moore Show the reasoning generated by the agent Question Is 702 interested in the area of expertise of 701 7 Answer Yes because 702 is interested in 703 which is the area of expertise of 701 Condition 701 is PhD advisor is expert in 703 702 is PhD student is interested in 703 703 is PhD research area 2008 Leaming Agents Center PhD research area subco ceptof Computer Science Artificial ntelli v enc Computer Networks PhD StUdent PhD advisor k Insta iceof insta CeOf isi Dan Moore Bill Jones Questions and Exercises Develop the ontology suggested by the following modeling We need to IAssess the student placement record of John Morris I I Who is a graduated PhD student of John Morris C Dan Adams who defended his PhD thesis in 2007 ITherefore we need to IAssess the reputation ofthe employer of Dan Adams I C What is the reputation of the employer of Dan Adams D I Dan Adams is employed by IBM which has the highest reputation 139 Therefore we conclude that I The reputation ofthe employer of Dan Adams is highest I 2008 Learning Agents Center Questions and Exercises Define the problem of inductive concept learning from examples What is abduction Give an example of abductive reasoning that was not discussed in class Provide two other explanations that are less plausible Specify a context when one of these alternative explanations would actually be more plausible See many questions and exercises in the lecture notes titled Overview of Basic Machine Learning Strategies 2008 Leaming Agents Center Questions and Exercises What is multistrategy learning What is the motivation of a multistrategy approach to learning What are the learning strategies used by Disciple during Rule Learning and Rule Refinement Answers Multistrategy learning denotes a type of learning that integrates several learning strategies such as empirical learning from examples explanationbased learning neural network learning or learning by analogy The singlestrategy learning methods have complementary strengths and weaknesses For instance empirical inductive learning from examples require many examples but does not require background knowledge In contrast explanation based learning requires only an example and complete background knowledge Multistrategy learning methods integrate different learning strategies to take advantage of their complementary strengths to compensate for their relative weaknesses The rule learning and refinement methods of Disciple integrate learning from examples learning from explanations and learning by analogy and experimentation 2008 Leaming Agents Center 45 Questions and Exercises Define the rule learning problem in Disciple Compare the rule learning process in the Disciple approach with the traditional knowledge acquisition approach where a knowledge engineer defines such a rule by interacting with a subject matter expert Identify as many similarities and differences as possible and justify the relative strengths and weaknesses of the two approaches but be as concise as possible Define the rule refinement problem in Disciple Compare the rule refinement process in the Disciple approach with the traditional knowledge acquisition approach Identify as many similarities and differences as possible and justify the relative strengths and weaknesses of the two approaches but be as concise as possible 2008 Leaming Agents Center 46 Questions and Exercises Consider the following example and associated explanation Task Assess the credibilith of EVD Davm Mjr l Dle Question How was EVDDawnMr lDle obtained Answer EVDDattm Pu rEIlUlo was obtained as testimonial evidence of Osama bin Laden cited in EVDDaum N r lUl by Hamid Mir Explanations Hamid Mir is a testimony by E39u D Dawn Mir l Ul is a testimony about 5 Eb39DDawn MirUI Elll is a testimony lay 25 Osama bin Laden g it a Subtask 3 Assess the eredib i 3 of Hamid Mir as the reporter of EVDDawnIt ljr 1 01 Sub task Assess the credibilityr of Osama bin Laden as the source of EVD DawnMirl 1 1I What rule will be learned from them assuming the ontology from the next two slides 2008 Leaming Agents Center Questions and Exercises is testimony by domain evidence range source sUaconcep of is testimony about domain evidence terrorist range eVIdence A instance of instar ce of Ham d Mir lOsama bin Laden 2008 Leaming Agents Center 48 Questions and Exercises lobject subconcept of evidence testimonial evidence subconcept of direct testimonial piece of evndence evidence 5 concept 0 subco cept of testimonial evidence obtained at second hand based on direct observation nonelementary elementary piec e testimonial evidence pieceofevidence ofevidence interview ance 039 EVDDawnMir O101 EVDDawnMir O1O1 c 2008 Leaming Agents Center 49 Questions and Exercises Consider the following example and associated explanation Example 1 of a problem reduction step We need to Which is a member ofAIIiedForces1943 has as member AlliedForces1943 us1943 us1943 Therefore we need to Camilla What rule will be learned from them assuming the ontology from the next slide 2008 Leaming Agents Center 50 Questions and Exercises subconcept of hasasmember instance of domain multimemberforce other feature gt I I I multi group multi state single state single group illegal force force force force group I multi state multi state I alliance coalition 39 dominaant partner equal partners multi sliiate alliance multi state alliance v European Axis 1943 Allied Forces 1943 em er Germany1943 2008 Leaming Agents Center Questions and Exercises Consider the additional positive example We need to Therefore we need to Indicate the refined rule 2008 Leaming Agents Center
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'