A Statistical Method for Translation Quality Assessment
Foreign Affairs College, Beijing
This paper attempts to introduce a new statistical method for making a quantitative assessment of the quality of a translation. A set of criteria was defined, against which the quality of a given unit of translation can be measured. These criteria provide a framework within which a mathematical model is constructed. It takes the fuzzy subset theory as its theoretical basis, trying to solve the problem of translation quality assessment by an approach which might be called analysis-synthesis fuzzy evaluation. Mathematical operations can be performed with the aid of a computer program to yield a grade of membership which indicates the quality of a given translation. The model proposed in this paper is, therefore, of methodological significance.
Up until now, translation quality assessment has remained qualitative rather than quantitative. Following this approach, what we do is give a [ p. 44 ]description of the quality of a work of translation in impressionistic terms in relation to certain standards. Expressions like Excellent, Good, Fair, and Poor are thus often used to indicate the quality of a translation. But these words are vague. An Excellent may not be that excellent, and a Poor may not be all that poor. These terms express values that are felt to be somewhat fuzzy. Therefore, quality of translation is usually considered to be something that cannot be assessed in a crisp manner. However, the fuzzy set theory as first advanced by Lotfi A. Zadeh in 1965 offers us a mathematical means with which translation quality assessment can be subjected to quantitative analysis.
Dubois, DidierHenri Prade
1980Fuzzy Sets and Systems: Theory and Applica-tions. New York: Academic Press.