Abstract
The large amount of digital information that is increasingly available in the manufacturing process makes information retrieval (IR) a critical issue in this knowledge-based manufacturing era. Many IR models including the vector space model (VSM), Boolean model, fuzzy set model and probability model have been proposed in the literature. However, the performance based on these models is not satisfactory to the expectations of the end users. The reasons contributing to the end users’ dissatisfaction are the imprecise query formulation and poor document representations. Most of the models of document representation are based on conventional statistical techniques. However, during manufacturing process document retrieval, the various qualitative data and attributes in the document database could not be easily analysed by using the current statistical approaches. This paper proposes a rough-set-based approach to enrich document representation. The document classification rules are generated and the premise terms are provided by the rough-set approach. Therefore, the retrieval performance of the VSM is enhanced through support from the rough-set-based approach. A case study that includes the comparison of manufacturing process document retrieval performance through the standard VSM and the rough-set-based approach is illustrated by empirical data. This paper forms the basis for solving many other similar document-retrieval problems that occur in manufacturing industry.