Granular Computing and Its Applications in Information Retrieval Support Systems

点击翻译按钮获取中文摘要

Granular computing explores multiple levels of granularity in thinking, problem solving, and information processing. It uses a hierarchical granular structure consisting of a family of integrative levels. A level consists of inter-connected and inter-acting granules of the same granularity or nature. A granule is a focal point used in an observation or a unit used in a discussion during problem solving, giving a local and partial description. The granules in a level collectively provide a global and complete description. A family of granular structures provides multiple views of the same problem. A central thesis of my research is a triarchic theory of granular computing encompassing the three perspectives of philosophical foundations, methodological foundations, and computational paradigms. It stresses a structured understanding of, and structured methods for solving, a real-world problem from multiple views and at multiple levels of abstraction within each view. The proposed research program aims to fully develop the triarchic theory by formulating and articulating granular computing ideas and to fully capitalize on them in the areas of human problem solving and intelligent systems. By examining and synthesizing results from existing studies, we plan to design a language for describing precisely multilevel granular structures, to formulate a set of general discipline-independent principles and mechanisms for understanding problem solving using granular structures, and to construct and compile a set of granular computing algorithms, including top-down, bottom-up, middle-out and progressive computing. An understanding of the underlying granule-based principles of human problem solving can be used to instruct and empower more people to be better at solving problems. The same principles can also be applied in designing intelligent systems with human-level intelligence. Concrete models, such as rough sets, formal concept analysis, and knowledge spaces, will be carefully investigated. Prototype information retrieval support systems will be implemented by exploring granular structures of individual documents, a document space, a user space, and a domain knowledge space.