Text mining with application to engineering diagnostics

Abstract

Our research focuses on text document mining with an application to engineering diagnostics. In automotive industry, the auto problem descriptions are often used as the first step of a diagnostic process that map the problem descriptions to diagnostic categories such as engine, transmission, electrical, brake, etc. This mapping of problem description to diagnostic categories is currently being done manually by mechanics that perform this task largely based on their memory and experience, which usually lead to lengthy repair processes, less accurate diagnostics and unnecessary part replacement. This paper presents our research in applying text mining technology to the automatic mapping of problem descriptions to the correct diagnostic categories. We present our results through the study on a number of important issues relating to text document classification including term weighting schemes, LSA and similarity functions. A text document categorization system is presented and it has been tested on a large test data collected from auto dealers. Its system performance is very satisfactory. © Springer-Verlag Berlin Heidelberg 2006.