Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....tools for modeling knowledge bases. For the clinical practice guidelines and protocols different representations have been proposed, as discussed in section 2. In the context of this paper a restricted first order logic representation embedded into the knowledge acquisition environment MOBAL [8] turned out to be a good choice. It fits particularly well because the protocol introduced in this paper is based on flow charts and propositional logic, which can easily be represented in first order logic. How well the approach applies to more complex protocol structures needs further ....

K.Morik, S.Wrobel, J.-U.Kietz, and W.Emde. Knowledge Acquisition and Machine Learning: Theory, Methods, and Applications. Academic Press, 1993.

....stocks in the wake of, possibly interleaved and implicit, deductive and inductive steps. For this reason, while [1] has been inspirational, the similarities of that work with the one reported here do not run deep. Another system that bears some resemblance to the work described here is Mobal [15]. Mobal can be seen as a knowledge acquisition environment that brings together several inductive logic programming schemes into an integrated whole and provides sophisticated services, such as theory restructuring, that the engine described in this paper is silent on. While the overall ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde, Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications, Academic Press, 1993.

....to select the correct ones from the generated ones. This system also maintains connections between rules and the examples they explain, making knowledge maintenance easier. Some work has also looked at interleaving machine learning and knowledge acquisition to make knowledge acquisition easier [45, 51, 38]. However, current systems still construct knowledge acquisition interfaces for contributors that need to be trained in using these tools. Out proposed work seeks to change this, enabling collection in plain English from untrained users. 2.5 NLP: text mining and question answering The advent of ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Ende. Knowledge Acquisition and Machine Learning - Theory, Methods and Applications. Academic Press, London, 1 edition, 1993.

....prove examples. It has been applied to single predicate learning in the system ICN. 1 Introduction In the field of Inductive Logic Programming (I.L.P. many works deal with the task of learning a definition of a concept from positive and negative examples of this concept and a knowledge base [15, 16, 10, 6, 11]. The predicates specified in the knowledge base are called the basic predicates, and the predicate that must be learned is called the target predicate. The goal is to learn a concept definition, which is complete, i.e. which proves that all the positive examples are true , and consistent, i.e. ....

.... logic programs, we have developed a general framework, based on a threevalued logic (ftrue, false, undefined g ) Few works, except as far as we know [1] use a three valued logic since usually, either a classical two valued logic, or a four valued logic (ftrue, false, unknown, inconsistent g ) [17, 11] is used. To handle general programs, an underlying problem is the choice of the semantics of negation. In Logic Programming, two three valued semantics are commonly used, Fitting semantics [3] and the well founded semantics [19] We present here Fitting semantics, but our framework can as well ....

[Article contains additional citation context not shown here]

K. Morik, S. Wrobel, J.U. Kietz, W. Emde, 1993. Knowledge Acquisition and Machine learning: Theory, Methods and Applications. Academic Press, London.

....perception integrating action features The feature p moving 7 characterizes a parallel movement along the perceived object. The fact p moving(t3,13,25,slowly,front,along door,right) e.g. represents a slow movement along a doorway, perceived by the right side sensors during the time interval [13,25] (Fig. 3, Pict. 1) Picture 2) shows a situation, in which the robot is standing in front of a wall after having perceived a movement along a doorway. This situation is represented by the fact standing(t25,25,29,in front of wall,right,large side, along door) standing 7 can be used as a memory ....

....cover all examples of a concept [6] These views have in common restricted first order logic as representation formalism. We apply rdt in the way described by Helft [6] and we only learn from positive examples 5 . The learner rdt is integrated in the knowledge acquisition tool mobal [13]. It uses a extended function free horn logic as representation formalism, also allowing negated literals. A special feature of rdt is the use of rule schemata, i.e. formulas with predicate variables instead of predicate symbols. They define sets of formulas of the same syntax. In this way, the ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning -- Theory, Methods, and Applications. Academic Press, London, 1993.

....for knowledge acquisition or workbenches for defining knowledge bases. A method that proved extremely beneficial for the knowledge acquisition task was the integration of knowledge acquisition with machine learning techniques [10] The drawback of these approaches, e.g. the work described in [5], however, was their rather strong focus on structured knowledge or data bases, from which they induced their rules. In contrast, in the Web environment that we encounter when building Web ontologies, the structured knowledge or data base is rather the exception than the norm. Hence, intelligent ....

.... incurring references from 20 completely different events journals: Domain Method Features used Prime purpose Papers Free Text Clustering Syntax Extract Buitelaar [3] Assadi [1] and Faure Nedellec [6] Inductive Logic Programming Syntax, Logic representation Extract Esposito et al. [5] Association rules Syntax, Tokens Extract Maedche Staab [14] Frequency based Syntax Prune Kietz et al. 13] Pattern Matching Extract Morin [15] Classification Syntax, Semantics Refine Schnattinger Hahn [8] Dictionary Information extraction Syntax Extract Hearst [9] Wilks [21] and ....

[Article contains additional citation context not shown here]

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge acquisition and machine learning: Theory, methods, and applications. Academic Press, London, 1993.

....base can be used without change with the identified problem solver, whether it can be transformed before use, or whether it is necessary to acquire a completely new knowledge base using a KA tool. Other work which has sought to provide a common framework for problem solving and KA includes MOBAL [15], VITAL [16] and NOOS [1] 2] Our contribution to the topic of knowledge base reuse is most closely related to Puppe s work [20] and the Protg project [8] Puppe has realised, as do O Hara et al. 14] that as experts learn more about the domain, their perspectives on the task may change, and ....

Morik, K., Wrobel, S., Kietz J-U., Emde, W., (1993), "Knowledge Acquisition and Machine Learning: Theory, Methods and Applications", Academic Press, London.

.... revision systems are found in (Shapiro, 1983; Sammut and Banerji, 1986; Pazzani, 1988; Muggleton and Buntine, 1988; Bergadano and Giordana, 1989; Duval, 1991; Wogulis, 1991; Richards and Mooney, 1991; Rouveirol, 1992; DeRaedt, 1992; Ndellec, 1992; Ndellec and Causse, 1992; Esposito et al. 1993; Morik et al. 1993; Feldman and Ndellec, 1994) The revision problem can be described as containing two problems, two phases, and two strategies. The two problems are the completeness problem: given a theory T, and an example E of the concept C, such that T does not recognize the positive example E of C as an ....

Morik K., Wrobel S., Kietz J. U., Emde W. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications, Academic Press, 1993.

....4 Least Generalizations and Greatest Specializations symbols are removed from clauses and put in the background knowledge by techniques suchas#attening #Rouveirol, 1992#. Well known ILP systems suchasFoil #Quinlan Cameron Jones, 1993#,Linus #Lavra#c D#zeroski, 1994# and Mobal #Morik, Wrobel, Kietz, Emde, 1993# all use only function free clauses. More than one half of the ILPsystems surveyed by Aha #1992# is restricted to function free clauses. Function free clauses are also su#cient for most applications concerning databases. Our second contribution shows that a set S need not have a least ....

Morik, K., Wrobel, S., Kietz, J.-U., & Emde, W. #1993#.Knowledge Acquisition and Machine Learning: Theory, Methods and Applications. Academic Press, London.

....prove examples. It has been applied to single predicate learning in the system ICN. 1 Introduction In the field of Inductive Logic Programming (I.L.P. many works deal with the task of learning a definition of a concept from positive and negative examples of this concept and a knowledge base [15, 16, 10, 6, 11]. The predicates specified in the knowledge base are called the basic predicates, and the predicate that must be learned is called the target predicate. The goal is to learn a concept definition, which is complete, i.e. which proves that all the positive examples are true , and consistent, i.e. ....

.... logic programs, we have developed a general framework, based on a threevalued logic (ftrue, false, undefined g ) Few works, except as far as we know [1] use a three valued logic since usually, either a classical two valued logic, or a four valued logic (ftrue, false, unknown, inconsistent g ) [17, 11] is used. To handle general programs, an underlying problem is the choice of the semantics of negation. In Logic Programming, two three valued semantics are commonly used, Fitting semantics [3] and the well founded semantics [19] We present here Fitting semantics, but our framework can as well be ....

[Article contains additional citation context not shown here]

K. Morik, S. Wrobel, J.U. Kietz, W. Emde, 1993. Knowledge Acquisition and Machine learning: Theory, Methods and Applications. Academic Press, London.

....L = fP (X) Q(X; a)g and the set of instantiations is I = f[P=t] Q=r; Q=s]g, the application of Sigma to L yields the substitution L Sigma = f[t(X) r(X; a) s(X; a) g. The concept of integrity constraints, as it will be used in the approach presented in this paper, has been adopted from [MWKE93]. An integrity constraint is a clause in which all variables are quantified, using both all quantifiers (8) and existence quantifiers (9) Those variables that occur in a positive atom, but not in a negated atom are existence quantified. All other variables are all quantified. Consider the ....

....of both the confirmation criterion and the pruning criterion, only have to be computed when testing a hypothesis. 4 Implementation and Experimental Results Icdt has been implemented in Quintus Prolog c fl . The integrity constraints discovery tool Icdt can be loaded into the system MOBAL [MWKE93] and operate on MOBAL s knowledge base. Predicate Arity jFactsj credit worthy 1 1002 drawing credit 1 1001 account type 2 1000 account balance 3 1000 credit 1 1000 Table 4. Some predicates of the model Account. Firstly, let us consider the model Account for credits of credit institutions. A ....

[Article contains additional citation context not shown here]

Katharina Morik, Stefan Wrobel, J.-U. Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993. No. 204.

....the search for hypotheses. Assuming ordered clauses to be given (no matter how the ordering has been achieved) Muggleton and Feng [16] define the depth and degree of their premise literals. By specifying maximal values on both, depth and degree, the hypothesis language is restricted. Morik et.al. [12] sort the premise literals of a rule in order to prune the search in the hypothesis space. They define the relation P between premise literals via the minimum distance of the variables occurring in the literals. But, given the rule C body L 1 ; L 2 ; L 3 ; L 4 p 3 (Tr; S; X;Y ) a(Tr; O; S; ....

K. Morik, St. Wrobel, J. U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory, Methods, and Applications. Addison Wesley, 1993.

....instead of building the hypothesis one clause at a time. Another system that follows this approach is TRACY by Bergadano and Gunetti [2] During the preprocessing phase of the background knowledge, Lime automatically extracts type and mode information. Similar issues are addressed by Morik et al. [16] in their system MOBAL. 1.2 Outline of the Paper The outline of the paper is as follows. In Section 2, we introduce the noise model and the Bayesian framework employed in Lime. In Section 3, we describe the hypothesis language of Lime and discuss the notion of simple clauses in some detail. ....

....conducted, considerable improvement in performance can be achieved if mode information was available. Lime extracts mode information from the data which enables it to skip clauses that are not determinate. This process is also detailed in [13] Another system that addresses these issues is MOBAL [16]. 4.2 Removing Redundancy There are three ways in which redundancy is removed from the background knowledge. First, if a set of relations are equivalent then only one needs to be considered in the inductive process. For this purpose two relations are said to be equivalent if they consist of ....

K. Morik, S. Wrobel, J-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory, Methods and Applications. Academic Press, 1993.

....systems like MOBAL interleave construction and application of the theory. Contradictions arising from applying the theory are used to correct and refine it. The theory is built and refined in several cycles using different sources of knowledge as described by the sloppy modelling paradigm [Mo93]. The need for theory revision is obvious in such a setting, because contradicting information is likely to arise and to demand correction of the theory that has been built so far. Two types of contradicting information can occur. The theory can be overly general, so that false facts are provable, ....

Morik, K., S. Wrobel, J.--U. Kietz. Knowledge Acquisition and Machine Learning: Theory, Methods, and Applications.

....two approaches. This integration is expected to be synergistic in effect, the resulting combined approach being more effective than either of its components. However, although there have been case studies documenting successful applications of these integrated techniques (Buntine Stirling, 1991; Morik, Wrobel, Kietz, Emde, 1993; Nedellec, Correia, Ferreira, Costa, 1994; Webb, 1996) no previous research has provided comparative evaluation of the relative merits of integrated approaches as opposed to either constituent approach on its own. In particular, it has not been demonstrated that integration of machine learning ....

....induction 3 approach gives superior performance in real knowledge acquisition tasks to noninteractive induction and to knowledge acquisition by interview , the basis for this conclusion appears to be subjective judgment rather than experimental evaluation. The knowledge acquisition tool MOBAL (Morik et al. 1993), and its forerunner BLIP (Morik, 1987) has been applied to a wide variety of knowledge acquisition tasks. These systems use first order representations and they provide tools for specifying ontologies, learning rules from examples, revising rules using examples, and learning new predicates from ....

[Article contains additional citation context not shown here]

Morik, K., Wrobel, S., Kietz, J.-U., & Emde, W. (1993). Knowledge acquisition and machine learning: Theory, methods, and applications. London: Academic Press.

....TR idea, other existing systems still exhibit a large range of variation. In order to illustrate this breadth, we will exemplarily describe some interesting approaches briefly highlighting their specific features. Mobal is aimed at supporting knowledge acquisition and knowledge base maintenance [20, 36]. In Mobal rules may be accompanied by an exception set, i.e. a rule may only be applied to facts not contained in this set. Further, a confidence can be attributed to rules and facts. When revising the knowledge base the system tries to minimize the loss of confidence. 3 Contrary to most ....

....refinement and the multi strategy approach as competing with the logicbased theory revision 12 approach, but we propose to develop well founded logic based methods towards the more sophisticated hybrid approaches. First contributions along this line of research came from the MOBAL group [20]; Wrobel demonstrates the cooperation of different knowledge sources during knowledge acquisition and revision [35, 36, 34] and grounds his approach in belief change [26] a research area which could also yield the very basic theoretical framework for a comprehensive view on KB maintenance. In our ....

K. Morik, S. Wrobel, J. Kietz, and W. Emde, Knowledge Acquisition and Machine Learning: Theory, Methods and Applications, Academic Press, 1993.

....First theorem proving is for this purpose too powerful and we can infer dependencies by transitivity which is really simple. Second, we can find dependencies in relational databases, which can not be stored in the main memory as PROLOG assertions. In most others ILP learning systems like RDT, [Morik et al. 1993], functional dependencies can not be expressed. Systems, which are closer to ours, are empirically compared in section 4. 3 Discovering Data Dependencies In this section we present the algorithms to infer integrity constraints, unary inclusion dependencies and functional dependencies. For more ....

Morik, K., Wrobel, S., Kietz, J. U., and Emde, W. (1993). Knowledge Acquisition and Machine Learning Theory, Methods, and Applications. Academic Press

....door and enter the room. To execute this kind of commands, the robot must be able to classify the objects it perceives with its sensors, i.e. to assign them to concepts. Additionally, the system must be able to perform actions with these objects like moving through the doorway of the door found. [Morik and Rieger, 1993] have shown that this requires perceptual features and action features to be integrated, and perceptual features to be action oriented and action features to be perception oriented. They developed a representation for these operational concepts based on first order logic. To simplify the ....

....in restricted first order logic. Nevertheless, applying these algorithms to real systems like autonomous mobile robots is a another challenging task. In this paper, we will first present the representation of the perceptions of a mobile robot and a scenario used for learning, both developed by [Morik and Rieger, 1993]. In Section 3, we will describe three different learning algorithms of inductive logic programming, ILP, the research area of learning Horn clause programs. We will also describe the results we got from applying these learning algorithms to our learning tasks. In Section 4, the new algorithm grdt ....

[Article contains additional citation context not shown here]

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993). Knowledge Acquisition and Machine Learning -- Theory, Methods, and Applications. Academic Press, London.

....[ Muggleton and Feng, 1990 ] FOIL [ Quinlan, 1990 ] ITOU [ Rouveirol, 1992 ] the linkedness condition (CLINT [ De Raedt, 1992 ] Input output mode (MIS [ Shapiro, 1983 ] etc. We will also study how to take into account bias language such as Higher order schemata (MOBAL [ Morik et al. 1993 ] MILES CTL [ Tausend, 1994 ] Language Biases represented by grammars, Cohen, 1993 ] clause sets (TRACY, Bergadano and Gunetti, 1994 ] ffl Consistency of the newly generated hypotheses (CHVS) The hypotheses of CHVS that are not consistent with FHS (the set of failed hypotheses, see ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, 1993.

....use can be found in Sim and Rennels [49] and Visser and Bench Capon [58] 2. Knowledge acquisition Ontologies describe and structure the entities and relations that need to be acquired for the domain under consideration. Examples of this kind of ontology use are CUE ( 31] 30] and MOBAL ([40]) 3. Knowledge system design: Ontologies are reusable constructs in the design of knowledge systems because they can be used to represent the invariant assumptions underlying different knowledge bases in the same domain. As such, they can be considered as initial building blocks of the knowledge ....

K. Morik, S. Wrobel, J-U. Kietz and W. Emde, Knowledge acquisition and machine learning: theory, methods and applications, Knowledge-Based Systems (Academic Press Limited, London, United Kingdom 1993).

....dependencies by some kind of transitivity which is really simple, theorem proving which is too powerful for this purpose. Second, we can find dependencies in relational databases, which can not be stored in the main memory as PROLOG assertions. In most others ILP learning systems like RDT, cf. [Morik et al. 1993], functional dependencies can not be expressed. Systems, which are closer to ours, are empirically compared in section 4. 2 Terminology and Data Dependencies Familiarity is assumed with definitions of relational database theory as given for example in [Kanellakis, 1990] The uppercase letters A; ....

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993). Knowledge Acquisition and Machine Learning: Theory, Methods and Applications. KnowledgeBased Systems. Academic Press, London u.a.

....(these are enhancements of Cobweb [27] and Protos [8] that among other improvements, can handle continuous valued attributes) At the micro level, each of these programs used several learning strategies to accomplish its subtask. Other examples of multistrategy systems are MOBAL (micro and macro, [59]) MLT (macro, 42] and MCS (micro, 13] ML techniques can be viewed not only as knowledge generation tools but more generally, as data analysis or information modeling tools similar to traditional statistical techniques. Both statistical and ML techniques can be viewed as approximating ....

Morik, K., Wrobel, S., Kietz, J. U., and Emde, W. Knowledge Acquisition and Machine Learning --- Theory, Methods and Applications. Academic Press, London, UK, 1993.

....the domains presented in [Tor95] the effects of language biases appeared as very extreme. Tor95] suggest some strategy to exploit this observation. 3.2. 3 A scheme based language for language bias The basic idea of the representation MILES CTL [Tau94a] is an extension of the approach in Mobal [MWKE94], i.e. sets of hypotheses are described by schemes. The aim for developping MILES CTL is to achieve an empirical comparison of the biases used in ILP in terms of the size of the hypothesis space. One major prerequisite was to extend the model of inductive learning in order to reveal the different ....

K. Morik, S. Wrobel, J. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, 1994.

....In view of the differing and complementary capabilities of knowledge elicitation and machine learning, there is considerable potential for gain through the integration of the two approaches. A number of research groups have investigated the integration of machine learning and knowledge elicitation [6, 7, 8, 9]. These approaches provide environments suited to sophisticated knowledge engineers. In contrast, the current research investigates the integration of knowledge elicitation with machine learning for users with little or no expertise in knowledge acquisition or knowledge based systems [10] In ....

K. Morik, S. Wrobel, J.-U. Kietz and W. Emde, Knowledge Acquisition and Machine Learning: Theory, Methods, and Applications (Academic Press, London, 1993).

....necessarily cover all examples of a concept [5] These views have in common restricted first order logic as representation formalism. We apply rdt in the way described by Helft [5] and we only learn from positive examples 2 . The learner rdt is integrated in the knowledge acquisition tool mobal [12]. It uses a extended function free horn logic as representation formalism, also allowing negated literals. A special feature of rdt is the use of rule schemata, i.e. formulas with predicate variables instead of predicate symbols. They define sets of formulas of the same syntax. In this way, the ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning -- Theory, Methods, and Applications. Academic Press, London, 1993.

No context found.

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993) Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications, Academic Press, London.

No context found.

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993) Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications, Academic Press, London.

No context found.

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993) Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications, Academic Press, London.

....(2) B; H j= E (completeness of H) 3) 8e 2 E : B; H 6j= e (accuracy of H) 2. 1 MOBAL MOBAL is a workbench which allows users to easily enter facts and rules, detects inconsistencies in the knowledge base, and proposes minimal changes to facts and rules in order to make it consistent [10]. In addition to the support of users in building up a knowledge base, the rule discovery tool automatically learns rules from facts and adds the learned rules to the knowledge base. 2.1.1 The Rule Discovery Tool RDT For learning rules from facts, the Rule Discovery Tool RDT forms all possible ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London, 1993.

....have not yet been acknowledged as a scientific matter in its own right. Some efforts have been made in order to support the layout of the knowledge representation language by a system, define sets of defining predicates for a learning goal, and to automatically adjust the representation language [55], 85] However, there are additional tasks that are not yet supported by tools. Hence, it was our goal to provide a general tool for the preparation of data for relational learning. Here is a list of the data engineering tasks that we considered: Data conversion: Very often, the given data do ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London, 1993.

....with a specific tool. The appendix contains ffl a brief discussion of the principles behind the system ffl lists of parameters and syntax for reference ffl some miscellaneous tips ffl a guide to installing Mobal A book on the system is to appear from Academic Press in winter of 1993 [ MWKE93 ] 8 Part I Basics 2 A GUIDED TOUR OF MOBAL 9 2 A guided tour of Mobal Mobal is an environment for incremental modeling. Although there is no fixed sequence of steps the user has to take in working with the system, some prototypical ways of using the system can be sketched. Two extremes are ....

.... Krt has been applied successfully in a number of domains, including a medical domain developed in cooperation with ICS FORTH, and the abovementioned telecommunications domain developed in cooperation with Alcatel Alsthom Recherche, Paris [ SMAU93 ] Its theoretical background is sketched in [ MWKE93 ] and fully described in [ Wro93a ] 6.2 Invoking KRT Krt can be invoked in two different ways. From a facts window. Double clicking on a fact in the facts window brings up the fact operations menu. Click on the button labeled Knowl. Revis. to call knowledge revision on the chosen fact. Krt ....

[Article contains additional citation context not shown here]

Katharina. Morik, Stefan Wrobel, J.-U. Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993. To appear. REFERENCES 119

....contains ffl a description of the input syntax. ffl some miscellaneous tips, and ffl a guide to installing Mobal Further reading This documents describes the services offered by Mobal and ways of using them. If you want to know exactly how Mobal realizes these services, please consult [ MWKE93 ] which contains an introduction to knowledge acquisition and machine learning, describes the different tools, their algorithms and their theoretical background, and contains a chapter about applications of Mobal. Further detail can of course be found in our other publications (see publication ....

....a domain model using only facts and rules. Metapredicates are also used in Mobal s inductive rule learning component Rdt to restrict the hypothesis space (see Chapter 5) In the following, we describe the different representational constructs and knowledge sources in more detail (see also [ MWKE93 ] 4.1 Facts Facts are used to state relations, properties of objects and concept membership. Facts are function free literals that must not contain variables. The form of a positive fact is p(t 1 ; t n ) where p is a n ary predicate and t j is a constant or a number. The form of a ....

[Article contains additional citation context not shown here]

Katharina. Morik, Stefan Wrobel, J.-U. Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993.

....by an expert, represent the effects of substances in different dosages, relations between vital signs, and interrelations between different substances, and validate the knowledge on the basis of past patients data. The knowledge acquisition and validation was supported by the MOBAL system [24]. validation of recommended interventions: Given . the state of a patient described in qualitative terms, medical knowledge . a sequence of interventions, and . a current intervention, find the effects of the current intervention on the patient. The derivation of effects is made for each ....

....of the data acquisition process at the hospital and the resulting data set [13] A statistical method for data abstraction is described in section 5. The next section (6) shows, how we applied the support vector machine (SVM) to learn state action rules. A short introduction to the MOBAL system [24] and its representation of medical knowledge leads to the issue of validation which is presented in section 8. 4 Data acquisition and data set 4.1 Data acquisition Most variables are entered by hand at the bedside. For entities such as clinical observations, nursing procedures, therapeutic ....

[Article contains additional citation context not shown here]

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde, Knowledge Acquisition and Machine Learning -- Theory, Methods, and Applications (Academic Press, London, 1993).

....set notation to minimally specialize clauses. Since there are usually several choices of which clauses to specialize, the operator produces a set of possible minimal base revisions. In the ILP project, the MBR operator is used in the Krt specializing revision system [56, 55] a part of Mobal [25]) Closed world specialization The ILP project has also investigated the closed world specialization (CWS) operator [6] that uses non monotonically interpreted exception predicates instead of exception lists, and thus stays within the standard language of logic programming. This representation ....

Katharina Morik, Stefan Wrobel, J.-U. Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993.

....of knowledge acquisition was started by the BMFT: the FABEL project. 4 MOBAL is the successor of BLIP. It has been developed at the German National Research Center for Computer Science in the course of the European project Machine Learning Toolbox (P2154) A detailed description of MOBAL is [Morik et al. 1993b] Machine learning does not replace a knowledge engineer in modeling. Instead, inductive algorithms in concert with other techniques are capable of assisting the knowledge engineer. 2 Logic oriented machine learning Predicate logic allows to write statements that are easier to understand than ....

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993b). Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London. to appear.

....databases without reducing it to the expressiveness of an algorithm such as Kid3, for instance. If we want to keep the capability of relational learning but also want to learn from all tuples of a large database, we need more restrictions. They 3 A detailed description can be found in (Morik et al. 1993). 4 Since Stt took about 8 hours, it cannot be subsumed under the fast algorithms. However, its result is computed only once from the background material which otherwise would have been ignored. should lead to a reduction of the number p of predicates or the maximal number i of attribute values ....

Morik, K.; Wrobel, S.; Kietz, J.-U.; and Emde, W. 1993. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. London: Academic Press.

....are deleted) Lavrac Dzeroski 94, p. 162 172] and the concept description becomes available to the overall system as well as to following learning processes. 3 Learning in relational domains Cola is an inductive learning tool in the knowledge acquisition and machine learning system Mobal [Morik et al. 93] and makes use of Mobal s knowledge representation environment. This means that Cola uses an extended function free Horn clause representation (paraconsistent with negation) Morik et al. 93, p. 27ff] The attributes of objects and relations among objects in the domain are described by facts. A ....

....domains Cola is an inductive learning tool in the knowledge acquisition and machine learning system Mobal [Morik et al. 93] and makes use of Mobal s knowledge representation environment. This means that Cola uses an extended function free Horn clause representation (paraconsistent with negation) Morik et al. 93, p. 27ff] The attributes of objects and relations among objects in the domain are described by facts. A set of facts related to an edge (named a2) in a finite element mesh model [Dolsak Muggelton 92] of a cylinder is shown as an example in table 1. All argunot important(a2) fixed(a2) not ....

Katharina Morik, Stefan Wrobel, Jorg-Uwe Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993.

....and is also independent from a particular conceptual clustering algorithm. The next section describes the details of an implemention of the method in a system called Cola. 4 Cola: The System Cola is an inductive learning tool in the knowledge acquisition and machine learning system Mobal [Morik et al. 93] and makes use of Mobal s knowledge representation environment. This means that Cola uses an extended functionfree Horn clause representation (paraconsistent with negation) Morik et al. 93, p. 27ff] tower1 tower2 tower3 arch1 arch2 wall1 wall2 Fig. 2. Various block world objects The attributes ....

....System Cola is an inductive learning tool in the knowledge acquisition and machine learning system Mobal [Morik et al. 93] and makes use of Mobal s knowledge representation environment. This means that Cola uses an extended functionfree Horn clause representation (paraconsistent with negation) Morik et al. 93, p. 27ff] tower1 tower2 tower3 arch1 arch2 wall1 wall2 Fig. 2. Various block world objects The attributes of objects and relations among objects in the domain are described by facts. A set of facts related to one building in a blocks world domain (see figure 2) is shown as an example in table ....

Katharina Morik, Stefan Wrobel, Jorg-Uwe Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993.

....is used when doing breadth first search for learning. Breadth first search allows to safely prune branches of sets of hypotheses that already have too few support in order to be accepted. Figures 1 and 2 show the pseudocode skeleton of the Rdt db algorithm, which remained the same as in Rdt (cf. [17][Chap. 6] RS and LEAVES represent the search status. At the beginning, RS contains all rule models, which conclusions are unifiable with Q. These rule models will be instantiated and tested breadth first according to RS ( subsumption defined on rule models) The data structure LEAVES stores all ....

....around. In this case, the first set of values is a subset of the second one. We omit the presentation of other features of Stt. Here, we apply it as a fast tool for finding sets of entities that are described by several predicates in background knowledge. 6 A detailed description can be found in [17]. We have represented textual background material as one ary ground facts. The predicates express independent aspects of attribute values of an attribute of the database. These attribute values are at argument position. Different predicates hold for the same attribute value. For 738 predicates ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London, 1993.

....revisions of theories in an incremental learning system. We then present the Mbr (for minimal base revision ) operator for revision of clausal first order theories that is shown to meet the revised set of postulates, and is fully implemented in Krt, the knowledge revision tool of the Mobal system [MWKEss]. It has been successfully used for revision of a set of telecommunication access control rules when faced with a number of incorrect access assignments in an application developed in cooperation with Alcatel Alsthom Recherche, Paris [SMAU93] In contrast to algorithms previously proposed by ....

....revision operations, we will now present a revision operator for clausal theories that meets all of these postulates. This operator, which we will call Mbr, for minimal base revision, and denote by Gamma, has been implemented and is used in Krt, the knowledge revision tool of the Mobal system [MWKEss], to perform knowledge revision. In order to define Gamma, we need to define the derivation tree of a fact in a theory. As usual, we will assume that is implemented as refutation proofs by resolution. Definition 6 (Derivation) Let Gamma be a theory, and f 2 Cn( Gamma) a factual query ....

K. Morik, S. Wrobel, J.U. Kietz, and Werner Emde. Knowledge Acquisition and Machine Learning: Theory, Methods, and Applications. Academic Press, London, New York, in press. To appear.

....GUI MASY ADLER Cobweb GPRA EVBS VWS KIS BWV ConceptBase Meta Data ADLER Figure 1: Architecture and Embedding of ADLER and a high extensibility. The heart of the DAWAMI architecture is the Data Mining workbench Kepler [WWSE96] This system is the successor of the machine learning toolbox Mobal [MWKE93] and relies on the idea of Plug Ins : Its tool description language and the primitives of a basic API enable the building of wrappers around a given implemented mining algorithm and its inclusion into the applicable set of tools. Figure 1 shows some examples for algorithms that are considered as ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory, Methods and Applications. Academic Press, 1993.

....as second argument of a predicate p 1 also occur as first argument of predicate p 3 , but not the other way around. In this case, the class class3 : ext(arg 2(p 1 ) becomes a subclass of class2 : ext(arg 1(p 3 ) We omit a detailed and formal presentation of Stt here and refer to chapter 4 in (Morik et al. 1993). The point here is that Stt can effectively group together attribute values on the basis of background knowledge. 17 We have represented textual background material about Mercedes car parts as unary ground facts. Stt forms classes on the basis of the given facts. The resulting semilattice of ....

Morik, K., Wrobel, S., Kietz, J.-U., & Emde, W. (1993). Knowledge acquisition and machine learning: Theory, methods, and applications. London: Academic Press.

....k nearest neighbor learning. The following subsections describe these modules, but first let us describe the knowledge representation we are using. 3. 1 Representation of the learning input Ribl is implemented as an external tool of the knowledge acquisition and machine learning system Mobal [ Morik et al. 1993; Sommer et al. 1994a ] Ribl induces concepts from the knowledge represented in the knowledge base of Mobal using an extended functionfree Horn clause representation that is para consistent with negation [ Wrobel, 1994 ] The concepts induced by Ribl are stored within Ribl, whereas the ....

K. Morik, S. Wrobel, J.-U. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, London, New York, 1993.

....of the robot towards the sensed edge. 2 MOBAL is developed at the German National Research Center for Computer Science. 3 Formal properties of the formalism have been proved in [Wrobel, 1987] A description of the system including detailed chapters about the representation formalism is [Morik et al. 1993]. 7 Rules We may view the rules in MOBAL as Horn clauses. If the premises can be instantiated by positive facts, the conclusion is derived. An example for a rule is the following stable(Trace,SAlpha,S,T1,T2,Value1) incr peak(Trace,SAlpha,S,T2,T3,Value2) stable(Trace,SAlpha,S,T3,T4,Value3) ....

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. (1993). Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London. to appear.

No context found.

Morik, K., Wrobel, S., Kietz, J.-U., and Emde, W. Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London, 1993.

No context found.

Morik, K., Wrobel, S., Kietz, J.U., Emde, W.: Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London (1993)

No context found.

K. Morik, S. Wrobel, J. Kietz, and W. Emde. Knowledge Acquisition and Machine Learning: Theory Methods and Applications. Academic Press, 1993.

No context found.

Morik, K., Wrobel, S., Kietz, J.U., Emde, W.: Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London (1993)

No context found.

Morik, K., Wrobel, S., Kietz, J.U., Emde, W.: Knowledge Acquisition and Machine Learning - Theory, Methods, and Applications. Academic Press, London (1993)

No context found.

K. Morik, S. Wrobel, J.U. Kietz, W. Emde, 1993. Knowledge Acquisition and Machine learning: Theory, Methods and Applications. Academic Press, London.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC