Skip to Main Content
Document databases may be ill-formed, containing redundant and poorly organized documents. For example, a database of customers’ descriptions of problems with products and the vendor's descriptions of their resolution may contain many descriptions of the same problem. A highly desirable goal is to transform the database into a concise set of summarized reports—model cases—which in turn are more amenable to search and problem resolution without expert intervention. In this paper, we describe techniques for attempting to automate the procedures for reducing a database to its essential components. Our initial application is self help for resolution of product problems. A lightweight document clustering method is described that operates in high dimensionality, processing tens of thousands of documents and grouping them into several thousand clusters. Techniques are described for summarization and exemplar selection to further refine the database contents. The method has been evaluated on a database of over 100,000 customer-service problem reports that are reduced to 3000 clusters and 5000 exemplar documents. Preliminary results are promising and demonstrate efficient clustering performance with excellent group similarity measures, reducing the original database size by several orders of magnitude.
Note: The Institute of Electrical and Electronics Engineers, Incorporated is distributing this Article with permission of the International Business Machines Corporation (IBM) who is the exclusive owner. The recipient of this Article may not assign, sublicense, lease, rent or otherwise transfer, reproduce, prepare derivative works, publicly display or perform, or distribute the Article.