An Analysis of Translation Quality in Multilingual Contexts

1. Introduction

Managers who work internationally recognize that failure to take into account the specific characteristics of the local operating environments, notably the official language and cultural differences, can have costly adverse effects. The multilingual capability of any information or document system’s applications that are intended to contribute toward companies’ international strategies or, more generally, to help reduce global differences in economic development is a key factor in its potential success. Not only is the automatic translation of some source language into one or more local languages and back extremely complex and costly, but linguistic usage may also be affected for practical reasons and, even more regrettably, in order to maximize or maintain the status of some language. The overall quality of the translation, including the naturalness of the style of the target language, is an important consideration and is somewhat independent of the translation system in use at the time the translation is done.

The purpose of this paper is to analyze fundamental aspects of translation quality measurement and to apply them to a specific multilingual context. A relationship between the quality perceived by users of translated documentation and selected translation measures has already been identified, but most impacts of the comparisons were identified only in isolation. Also, the selection of thresholds for stopping automatic translation was applied using language-independent measures without the threshold question being approached from a distance. We will describe the unsupervised development of a measure that, among language pairs, reflects desirable document translation performance results. The development takes into account the time for the individual agent within the context. These comparisons, in terms of the evolving uses of concepts, serve as one of the method requirements of multi-metric programs.