Knowledge discovery in the Internet

With the rapid expansion of the World Wide Web, the need for efficient data retrieval strategies becomes stronger and will be still growing. Unfortunately classical information retrieval techniques, developed for well-organized collections of textual data do not seem to be able to cope with diversity and amount of information available throughout the Internet. This paper presents some of the newest approaches to information retrieval in large, unstructured hypertext spaces such as WWW that focus more on latent information embedded in hyperlinks and document structure, then on actual understanding of Web pages textual content. These techniques, that are marking the new trends and prospects for the Internet technology, have been given recently the name "Web mining", as in fact they are examples of unsupervised machine learning similar to data mining and text mining. Here we discuss methods belonging to the following three groups: link topology analysis, statistical text analysis and query languages and systems design.