Corpus Building for Corporate Knowledge Discovery and Management: A Case Study of Manufacturing

摘要

Building a collection of electronic documents, e.g. corpus, is a cornerstone for the research in information retrieval, text mining and knowledge management. In literature, very few papers have discussed the necessary concerns for building a corpus and explained the building process systematically. In this paper, we explain our work of building an enterprise corpus called manufacturing corpus version 1 (MCV1) for corporate knowledge management purpose. Relevant issues, e.g. input texts, category labels and policies, as well as its parallel coding process and quality measurements are discussed. The real-world automated text classification experiments based on MCV1 show the soundness of its coding process. Finally, suggestions are made on how the proposed approach can be implemented in a more economical manner.