A study of granular computing for the organization and retrieval of XML documents

The common practice for the design and implementation of an information retrieval system is to use a single representation of documents and a single retrieval method. However, an information retrieval system deals with a diversity of users. The users may have different information needs and retrieval requirements. As well, there does not exist a unified retrieval method that can solve various retrieval problems. In addition, the users may also want to explore a document space at different levels of abstraction. An information retrieval system designed by using a single representation of documents and a single retrieval method cannot be adaptive to different retrieval situations.; Based on XML and granular computing, the present study proposes a granular computing model for the organization and retrieval of XML documents. The model is developed under the consideration of a multi-representation of documents, a multi strategy retrieval, and user involvement and support. An XML document collection can be conceptually represented by two matrixes, namely, a document-by-element-by-term matrix and a document-by-attribute matrix. Various granulated views of the collection can be generated based on the two matrixes. Different retrieval methods can be used under different granulated views. Different retrieval problems can be dealt with by combining different granulated views and different retrieval methods.; A granulation language is also developed to support the users in creating granulated views, choosing retrieval methods, and navigating a granulated document space. With the support of the language, the users may explore a document space based on the structural and semantic information at both the document level and the collection level.; The proposed model provides a framework for designing an information retrieval system that can be adaptive to various retrieval situations.