A query-based cross-language diagnosis tool for distributed decision making support

Abstract

A query translation-based Korean–English cross-language diagnosis (Q-KE-CLD) tool for assisting Korean users diagnosing print defects was developed and then evaluated as a case study of distributed decision making support for nonnative English users. The first step in developing the Q-KE-CLD tool involved collecting and analyzing print defect descriptions in Korean and English. A fuzzy Bayesian model was obtained from the descriptions and the Q-KE-CLD tool was developed. The tool was then experimentally evaluated in four different universities in South Korea. Results showed that Korean subjects generated Korean queries faster (p = 0.008) when entering Korean queries. In addition, the subjects rated Korean queries as being easier to generate (p = 0.004). Untrained subjects reported that use of the Korean language made it easier to generate queries and identify print defects. The overall results suggested that query translation-based cross-language diagnosis is a feasible approach for localizing troubleshooting websites.

Introduction

Complex organizations in general have a decision making structure in which distributed decision makers share information (Barber et al., 2000, Boland et al., 1992, Jennings, 1966). Computer technologies have played an important role in integrating distributed sources of information and improving access (Kraemer and King, 1988, Pinson et al., 1997). Internet technologies have especially accelerated development of distributed decision making systems in which customers perform tasks such as troubleshooting jointly with decision makers.

In recent years, many companies have developed websites that their customers can use to diagnose product problems (Choe et al., 2006, Foo et al., 2000, Kim et al., 2005). A good troubleshooting website can help minimize the need to maintain expensive call centers, and can improve the customer experience. However, most websites include text that users have to read, understand, and follow to navigate the websites. This can obviously cause problems for nonnative users, such as when a Korean user tries to navigate a website containing written English. The traditional solution has been to develop localized websites which present information to users in their native language (Aykin, 2005, Brandon and Jr., 2001). The most expensive and time consuming method of localization is to translate the entire website. An alternative approach proposed in this study focuses on allowing the users to enter queries in the website in their native language to diagnose their problems. One advantage of this approach is that query diagnosis quickly leads to target information minimizing the process of website navigation (Olston & Chi, 2003). In addition, queries are much easier to translate into other languages than entire websites, which means that query translation-based cross-language information retrieval (Q-CLIR) might be a more practical approach for localization than translating the website itself.

CLIR is defined as a method for finding relevant documents in one language (or target language) using a query expressed in another language (Chen, 2003, Diekema, 2003, Lee et al., 2004, Mitkov, 2003, Petrelli et al., 2002, Salton, 1970, Salton, 1973). Previous applications of Q-CLIR have focused for the most part on allowing users to retrieve information from large databases. It seems reasonable that this approach could be easily applied to help users diagnose or troubleshoot problems with a product in a website resulting in a method we call query translation-based cross-language diagnosis (Q-CLD). In such an approach, a user would diagnose product problems by entering a query describing the problem. The system would then translate the query and return potential diagnoses.

In this paper, we present a case study where a Q-CLD tool for query translation-based Korean–English cross-language diagnosis (Q-KE-CLD) for print quality troubleshooting was implemented and evaluated. The main objective of the research was to evaluate the Q-KE-CLD tool to see if a query translation-based cross-language diagnostic approach can reduce the time needed by users to diagnose problems, as well as increase user’s satisfaction. In addition to being of theoretical interest, this research was prompted by the pressing need of Hewlett Packard and other companies for localized print quality troubleshooting websites. Over the past 5 years, print quality troubleshooting websites for several lines of laser printers have been developed by Purdue University and Hewlett Packard (ex. http://www.hp.com/cpso-support-new/pq/5500/home.html). The websites allow users to diagnose print defects associated with several different models of Hewlett Packard’s laser printers by link-based browsing (Choe et al., 2006, Kim et al., 2007, Kim et al., 2005, Park et al., 2006). This existing website provided an excellent test bed for evaluating the potential benefit of the Q-CLD tool. This study therefore provided a practical approach to localize the print quality troubleshooting websites based on a Q-CLD.