MINING IS-PART-OF ASSOCIATION PATTERNS FROM SEMISTRUCTURED DATA
Ke Wang, Huiqing Liu · 2001
Introduction 1.1 Motivation As the amount of data available on-line grows rapidly, we find that more and more of the data are semistructured. Semistructured data arise when the source does not impose a rigid structure such as the Web and when data are combined from several heterogeneous sources such as a data warehouse. See, for example, 1;6;15 for motivation and the workshop 20 for some recent works. Semistructured data have two essential features: Structure. Semistructured data do have structures. Figure 1 shows a segment of semistructured movie objects maintained by IMDb (http://us.imdb.com). Each circle plus the text inside it represents an object, e.g., a HTML subdocument, and its identifier, e.g., the URL. The arrows and their labels represent object references and their roles. Object references can be cyclic, ordered, and up to an arbitrary level. Object references and their la