In order to cope with the high degree of freedom immanent in Internet data, the use of sensible categories to present a clear basis for machine learning systems, such as special sets of words and phrases (training classifiers) is needed. Also, training classifiers are substantial within the task of automated spam filtering
Published in Chapter:
Web Mining: A Conceptual Overview on Intelligent Information Retrieval Systems
Henrik Hanke (University of Duisburg-Essen, Germany) and Alf Neumann (University of Cologne, Germany)
Copyright: © 2008
|Pages: 6
DOI: 10.4018/978-1-59140-993-9.ch096
Abstract
The scale and scope of information on the Internet has been extended enormously over the past decade. The growth of more and more intelligent Web-based services and applications has resulted in an enormous growth of potentially useful data of both commercial and non-commercial interest. While this rise has brought a great amount of positive impact on global economic, social, and political development, it also implies an enormous flood of information into an increasingly complex information space. This is to be found on a vast variety of topics originating from a vast variety of sources, which range from private Web sites containing different kinds of information, to business-to-business B2B platforms. In most cases, these data are of an unsorted and unstructured kind, making efficient and target-oriented information retrieval very hard, if not nearly impossible. Coping with the challenge of a lack of transparency can be remedied by intelligent software agents, also referred to as softbots, which guide users through finding, sorting, and filtering this accruing data on the Internet like commonly used search engines.