In many text mining applications, knowledge bases incorporating expert knowledge are beneficial for intelligent decision making. Refining an existing knowledge base from a source domain to a different target domain solving the same task would greatly reduce the effort required for preparing labeled training data in constructing a new knowledge base. We investigate a new framework of refining a kind of logic knowledge base known as Markov Logic Networks (MLN). One characteristic of this adaptation problem is that since the data distributions of the two domains are different, there should be different tailor-made MLNs for each domain. On the other hand, the two knowledge bases should share certain amount of similarities due to the same goal. We investigate the refinement in two situations, namely, using unlabeled target domain data, and using limited amount of labeled target domain data. / When manual annotation of a limited amount of target domain data is possible, we exploit how to actively select the data for annotation and develop two active learning approaches. The first approach is a pool-based active learning approach taking into account of the differences between the source and the target domains. A theoretical analysis on the sampling bound of the approach is conducted to demonstrate that informative data can be actively selected. The second approach is an error-driven approach that is designed to provide estimated labels for the target domain and hence the quality of the logic formulae captured for the target domain can be improved. An error analysis on the cluster-based active learning approach is presented. We have conducted extensive experiments on two different text mining tasks, namely, pronoun resolution and segmentation of citation records, showing consistent ii improvements in both situations of using unlabeled target domain data, and with a limited amount of labeled target domain data. / When there is no manual label given for the target domain data, we re-fine an existing MLN via two components. The first component is the logic formula weight adaptation that jointly maximizes the likelihood of the observations of the target domain unlabeled data and considers the differences between the two domains. Two approaches are designed to capture the differences between the two domains. One approach is to analyze the distribution divergence between the two domains and the other approach is to incorporate a penalized degree of difference. The second component is logic formula refinement where logic formulae specific to the target domain are discovered to further capture the characteristics of the target domain. / Chan, Ki Cecia. / Adviser: Wai Lam. / Source: Dissertation Abstracts International, Volume: 73-02, Section: B, page: . / Thesis (Ph.D.)--Chinese University of Hong Kong, 2010. / Includes bibliographical references (leaves 120-128). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [201-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract also in Chinese.

Data mining algorithms discover knowledge from data. The knowledge are commonly expressed as dependency relationships in various forms, like rules, decision trees and Bayesian Networks (BNs). Moreover, many real-world problems are multi-class problems, in which more than one of the variables in the data set are considered as classes. However, most of the rule learners available were proposed for single-class problems only and would produce cyclic rules if they are applied to multi-class ones. In addition, most of them produce rules with conflicts, i.e. more than one of the rules classify the same data items and different rules have different predictions. Similarly, existing decision trees learners cannot handle multi-class problems, and thus cannot detect and avoid cycles. In contrast, BNs represent acyclic dependency relationships among variables, but they can handle discrete values only. They cannot manage continuous, interval and ordinal values and cannot represent higher-order relationships. Consequently, BNs have higher network complexity and lower understandability when they are used for such problems. / This thesis has studied in depth discovering dependency relationships in various forms by Evolutionary Computation (EC). Through analysis of the reasons leading to the disadvantages of rules, decision trees and BNs, and their learners, we have proposed a sequence of EAs, a novel functional dependency network (FDN) and two techniques for dependency relationship learning and for multi-class problems. They are the multi-population Genetic Programming (GP) using backward chaining procedure and the GP employing co-operating scoring stage for acyclic rules learning. The dependency network with functions can manage all kinds of values and represent any kind of relationships among variables, the flexible and robust MDLGP to learn the novel dependency network and BN. Based on the FDN we have further developed the techniques to learn rules without conflict and acyclic decision trees for multi-class problems respectively. The new self-organizing map (SOM) with expanding force for clustering and data visualization for data preprocessing have also been given in the appendix. / Shum Wing Ho. / "May 2007." / Adviser: Kwong-Sak Leung. / Source: Dissertation Abstracts International, Volume: 69-01, Section: B, page: 0436. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2007. / Includes bibliographical references (p. 221-240). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

An extension of wrapper adaptation is developed to collectively extract information from multiple Web pages. There exists mutual influence between text fragments of different Web pages and hence they should be considered collectively during extraction. Extending from the dependence model, a framework which can consider the dependence between text fragments within a single Web page and the dependence between text fragments from different pages. One characteristic of this model is that additional information can be incorporated into the model and multiple tasks earl be tackled simultaneously. As a result, a global solution which can optimize the quality of the tasks, at the same time, eliminate the conflict between them can he obtained. Experiments on product feature extraction and hot item mining from multiple auction Web sites have been conducted to demonstrate the effectiveness of this framework. / One problem of most existing Web information extraction methods is that the extraction knowledge learned from a Web site can only be applied to Web pages from the same site. This thesis first investigates the problem of wrapper adaptation which aims at adapting a wrapper previously learned from a source site to new unseen sites. A dependence model that can model the dependence between text fragments in Web pages is developed. Under this model, two types of text related features are identified. The first type of features is called site invariant features. These features likely remain unchanged in Web pages from different sites in the same domain. The second type of features is called site dependent features. These features are different in Web pages collected from different Web sites, while they are similar in Web pages originated from the same site. Based on this model, two frameworks are developed to solve the wrapper adaptation problem. The first framework is called Information Extraction Knowledge Adaptation using Machine Learning approach (IEKA-ML). Machine learning methods are employed to derive site invariant features from the previously learned extraction knowledge and items previously collected or extracted from the source Web site. Both site dependent features and site invariant features in new sites are considered for learning of new information extraction knowledge tailored to the new unseen site. / The second framework, called Information Extraction Knowledge Adaptation using Bayesian learning approach (IEKA-BAYES), solves the problem of wrapper adaptation as well as the issue of new attribute discovery. The new attribute discovery problem aims at extracting new or previously unwell attributes that are not specified in the wrapper. To harness the uncertainty, a probabilistic generative model for the generation of text fragments and layout format related to attributes in Web pages is designed. Bayesian learning and expectation-maximization (EM) techniques are developed under the proposed generative model to accomplish the wrapper adaptation task. Previously unseen attributes together with their semantic labels earl be discovered via another EM-based Bayesian learning on the generative model. Extensive experiments on over 30 real-world Web sites in three different domains and comparison between existing works have been conducted to evaluate the IEKA-ML and IEKA-BAYES frameworks. / Wong Tak Lam. / "October 2006." / Adviser: Lam Wai. / Source: Dissertation Abstracts International, Volume: 68-09, Section: B, page: 6095. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2006. / Includes bibliographical references (p. 126-135). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstract in English and Chinese. / School code: 1307.

Typescript (photocopy). / Digitized by Kansas Correctional Industries / Department: Computer Science.

New Zealand�s main export earnings come from the primary production area including agriculture, horticulture, and viticulture. One of the major contributors in this area of horticulture is the production of quality export grade fruit; specifically apples. In order to maintain a competitive advantage, the systems and methods used to grow the fruit are constantly being refined and are increasingly based on data collected and analysed by both the orchardist who grows the produce and also researchers who refine the methods used to determine high levels of fruit quality. To support the task of data analysis and the resulting decision-making process it requires efficient and reliable tools. This thesis attempts to address this issue by applying the techniques of Connectionist-Based Intelligent Information Systems (CBIIS) for Image Analysis and Knowledge Discovery. Using advanced neurocomputing techniques and a novel knowledge engineering methodology, this thesis attempts to seek some solutions to a set of specific problems that exist within the horticultural domain. In particular it describes a methodology based on previous research into neuro-fuzzy systems for knowledge acquisition, manipulation, and extraction and furthers this area by introducing a novel and innovative knowledge-based architecture for knowledge-discovery using an on-line/real-time incremental learning system based on the Evolving Connectionist System (ECOS) paradigm known as the Evolving Fuzzy Neural Network (EFuNN). The emphases of this work highlights knowledge discovery from these data sets using a novel rule insertion and rule extraction method. The advantage of this method is that it can operate on data sets of limited sizes. This method can be used to validate the results produced by the EFuNN and also allow for greater insight into what aspects of the collected data contribute to the development of high quality produce.

Australian agriculture is one area in which a number of intelligent support systems have been developed. It appears, however, that comparatively few of these systems are widely used or have the impact the developers might have wished. In this study a possible explanation for this state of affairs was investigated. The development process for 66 systems was examined. Particular attention was paid to the nature of user involvement, if any, during development and the relationship to system success. The issue is not only whether there was user involvement but rather the nature of the involvement, that is, the degree of influence users had during development. The patterns identified in the analysis suggest user influence is an important contributor to the success of a system. These results have theoretical significance in that they add to knowledge of the role of the user in the development of intelligent support systems. The study has drawn together work from three areas: Rogers’ diffusion theory, the technology acceptance model, and theories relating to user involvement in the development of information systems. Most prior research in the information systems area has investigated one or two of the above three areas in any one study. The study synthesizes this knowledge through applying it to the field of intelligent support systems in Australian agriculture. The results have considerable practical significance, as apparently developers of intelligent support systems in Australian agriculture do not recognize the importance of user participation, and continue to develop systems with less than optimum impact.

Rule induction is a popular technique for knowledge acquisition and data mining. Many techniques, such as ID3, C4.5, CART (tree induction tecniques) and Artificial Neural Networks have been developed and widely used. However, most techniques are either based on categorical or numerical mechanisms to assess the importance of different input variables, which may not produce the optimal rule when a mixture of variables exists. In 1992, Liang proposed a composite approach called CRIS that use different method to analyze different types of data in inducing rules for binary classification. Yang conducted a follow-up research to extend the original algorithm to multiple categories. However, both methods do not take variable interaction into consideration. The purpose of this research is to extend previous approach and extend by including second-order interaction. We also take into consideration the kurtosis and skewness of data for numerical variables. For categorical data, we also adopt ID3 algorithm to handle classes with low representation in the sample. In order to evaluate this technique, we develop a prototype CRIS 3.0 and compare with existing techniques, including multi-category-CRIS, CART and C4.5 as benchmark. The results show that CRIS 3.0 has the highest probability of producing the highest prediction accuracy.

Thesis (M.S.)--West Virginia University, 2004. / Title from document title page. Document formatted into pages; contains xiii, 148 p. : ill. (some col.). Includes abstract. Includes bibliographical references (p. 90-95).

Thesis (Ph. D.)--University of Washington, 2003. / Vita. Includes bibliographical references (leaves 223-231).

Thesis (M.S.)--Worcester Polytechnic Institute. / Keywords: run-time monitoring; continual validation; software probes; probing. Includes bibliographical references (p. 98-101).