Query translation-based cross-language print defect diagnosis based on the fuzzy Bayesian model

This paper discusses a query-translation based cross-language diagnosis (Q-CLD) for print defects conducted by nonnative English users. The first step involved developing three fuzzy Bayesian models: one based on English descriptions provided by native English subjects (referred to as the native English model); the second on English des- criptions provided by Korean subjects (referred to as the nonnative English model); and the third on Korean descriptions provided by Korean subjects (referred to as the Korean model). Model performance was evaluated using five different types of input descriptions. The results showed that the keywords matching translations developed in this study were nearly as accurate as the native English descriptions which were the most accurate predictions of the tested models. Using the keywords matching translations, the native English model correctly predicted 37% of the print defects with its top prediction and, in 80% of the cases the actual defect was one of the top five predictions. Considering that the native English model correctly predicted 45% of the print defects with its top prediction, and in 87% of the cases the actual defect was one of the top five predictions, the result supported the idea that a Q-CLD could be a practical localization approach for a troubleshooting website.