Text mining for a clear picture of defect reports: A praxis report

We applied the text mining categorization technology, in the publicly available, IBM Enterprise Information Portal V8.1 to more than 15,000 customer reported, product problem records. We used a proven software quality category set to categorize these problem records into different areas of interest. Our intent was to develop a clear picture of potential areas for quality improvement in each of the software products reviewed, and to provide this information to development's management. The paper presents the benefits that can be gained from categorizing problem records, as well as the limitations. © 2003 IEEE.