William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....computed using Church s part of speech tagger (1988) This gives them a recall rate of 39.0 and a precision of 55.2 . Litman (1996) explored using machine learning techniques to automatically learn classification rules for discourse markers. She contrasted the performance of CGRENDEL (Cohen 1992; 1993) with C4.5 (Quinlan 1993) CGRENDEL is a learning algorithm that learns an ordered set of if then rules that map a condition to its mostlikely event (in this case discourse or sentential interpretation of potential discourse marker) C4.5 is a decision tree growing algorithm that learns a ....

Cohen, W. W. 1992. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning.

....for bias specification At present there exist four more or less general frameworks to specify language bias, i.e. to specify the set of clauses allowed in hypotheses. This includes: the inductive logic programming language of Bergadano [8, 10] the antecedent description grammars of Cohen [24, 23], the schemata of the BLIP MOBAL team [35, 54] and their variants [111, 127, 135] The fourth framework, parametric languages as defined by [90, 107, 20, 25] will be presented when discussing the link to the complexity of learning. Bergadano s inductive logic programming language uses a notation ....

W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the 9th International Conference on Machine Learning. Morgan Kaufmann, 1992.

....the data to construct rules that best predicted the classifications from the features. This paper examines the utility of machine learning for automating the construction of rules for classifying cue phrases. A set of experiments are conducted that use two machine learning programs, cgrendel (Cohen 1992; 1993) and C4.5 (Quinlan 1986; 1987) to induce classification rules from sets of preclassified cue phrases and their features. To support a quantitative and comparative evaluation of the automated and manual approaches, both the error rates and the content of the manually derived and learned ....

....class in the corpus (sentential) has an error rate of 39 and 41 for the classifiable tokens and the classifiable nonconjuncts, respectively. Experiments using Machine Induction This section describes experiments that use the machine learning programs C4.5 (Quinlan 1986; 1987) and cgrendel (Cohen 1992; 1993) to automatically in duce cue phrase classification rules from both the data of (Hirschberg Litman 1993) and an extension of this data. cgrendel and C4.5 are similar to each other and to other learning methods (e.g. neural networks) in that they induce rules from preclassified ....

[Article contains additional citation context not shown here]

Cohen, W. W. 1992. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning. Aberdeen, Scotland: Morgan Kaufmann.

.... representation for concepts and examples [Cohen and Hirsh, 1992; Vilain et al. 1990; Kietz and Morik, 1991] most researchers have used standard first order logic as a representation language; in particular, most have used restricted subsets of the Prolog programming language to represent concepts [Cohen, 1992; Muggleton and Feng, 1992; Pazzani and Kibler, 1992; Quinlan, 1990; Muggleton, 1992a] The term inductive logic programming (ILP) has been used to describe this growing body of research. One advantage of basing learning systems on Prolog is that its semantics and complexity are mathematically ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

....learn from positive examples only; ffl the ability to handle high arity relations; ffl the ability to make use of domain specific background knowledge. The learning system we choose for further experimentation in the view specification recovery domain is Grendel2, a successor system to Grendel [Cohen, 1992]. Grendel2 employs a declarative bias. A declarative bias means that in addition to the usual set of positive and negative examples, the user can also provide to Grendel2 an explicit description of the space of possible hypotheses of the learning system. This bias description can alternatively ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

.... description logics as a representation for concepts and examples [Vilain et al. 1990; Kietz and Morik, 1991; Cohen and Hirsh, 1992] most researchers have used standard first order logic as a representation language; in particular, most have used restricted subsets of Prolog to represent concepts [Cohen, 1992; Muggleton and Feng, 1992; Pazzani and Kibler, 1992; Quinlan, 1990; Muggleton, 1992c] The term inductive logic programming (ILP) has been used to describe this growing body of research. One advantage of basing learning systems on Prolog is that its semantics and complexity are mathematically ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

.... systems on many problems [ Pagallo and Hassler, 1990; Quinlan, 1987; Weiss and Indurkhya, 1991 ] Finally, many sorts of prior knowledge about a learning problem can be communicated to a rule induction system by providing appropriate constraints on the form of induced rules [ Cohen, 1991; Cohen, 1992a ] The goal of this paper is to study the degree to which rule set induction methods scale up to large, real world learning problems. In particular, we will study the asymptotic complexity of rule induction in the presence of two phenomena common in real world datasets: large training set size ....

....runtimes still needs to be verified experimentally. The MDL based prepruning, for example, may need modification to work properly on relational problems. Also, while the techniques described here have natural extensions which can be applied in learning from examples and a grammatical bias [ Cohen, 1992a ] these extensions have again not been experimented with systematically. Acknowledgments Many thanks to Jason Catlett, Haym Hirsh, and Mike Pazzani for comments on an early draft of this paper, to Susan Cohen for proofreading it, and to Wan Ping Chiang for providing the rds data. ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

....to support inductive reasoning i.e. learning. Our analysis has focused on determining which description logics are learnable in Valiant s [ 1984 ] model of pac learnability [ Cohen and Hirsh, 1992b ] and with understanding the complexity of the operations necessary to support learning [ Cohen et al. 1992 ] Also at Computer Science Department, Rutgers University, New Brunswick, NJ 08903. In this paper, we build on these formal results in several ways. We extend the previous formal results to the description logic C Classic, which is expressive enough to be practically useful. We also present ....

....subsumes all of the D i s. An LCS can also be thought of as the dual of the intersection (AND) operator, or as encoding the largest expressible set of commonalities between a set of descriptions. A fairly general method for implementing LCS algorithms is described by Cohen, Borgida and Hirsh [ Cohen et al. 1992 ] This method can be used to derive an LCS algorithm for any description logic which uses structural subsumption : this class includes CClassic, but not full Classic2. An LCS algorithm for C Classic description trees is shown in Figure 2. 2 This algorithm produces a unique LCS for any set of ....

[Article contains additional citation context not shown here]

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

....in fact, some of the biases suggested by formal analysis have not (up until now) been implemented. A rapid prototyping system is an ideal mechanism for exploring this space. In this paper, we will describe such a system. Our prototyping system is based on a learning system called Grendel [ Cohen, 1992 ] In Grendel the user is provided with a language to express his or her choice of a bias; this language is called the bias representation language (BRL) More specifically, Grendel accepts two inputs: a set of examples of the concept to be learned, and an explicit description of the hypothesis ....

.... pred2(X,Y) adj(X,Y) In Grendel, ADG s are used to generate clauses. The space of possible clauses defined by an ADG is searched using a top down, separate and conquer greedy search procedure similar to that used by FOIL [ Quinlan, 1990 ] The search procedure is discussed briefly in [ Cohen, 1992 ] and at length in [ Cohen, 1991 ] 2 More precisely the non deferred portion of the grammar is used unless no clauses derived by these rules (and visible to the search procedure) have positive information gain. In addition, Grendel s search procedure imposes a preference bias towards smaller ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

....introduced by Pitt and Warmuth [ 1990 ] as a consequence our results are independent of the representations used by the learning system. Introduction Recently, there has been an increasing amount of research in learning restricted logic programs, or inductive logic programming (ILP) Cohen, 1992; Muggleton and Feng, 1992; Quinlan, 1990; Muggleton, 1992a ] One advantage of using logic programs (rather than alternative first order logic formalisms [ Cohen and Hirsh, 1992 ] is that its semantics and computational complexity have been well studied; this offers some hope that learning ....

William W. Cohen. Compiling knowledge into an explicit bias. In Proceedings of the Ninth International Conference on Machine Learning, Aberdeen, Scotland, 1992. Morgan Kaufmann.

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