An argumentation-centred approach to translation quality assessment.

Translation quality assessment (TQA) types may be divided into two main types: (1) models with a quantitative dimension, such as SEPT (1979), Sical (1986) and GTS (2000), and (2) non-quantitative, more textological models such as Nord (1991) and House (1997). Because it tends to focus on microtextual (lexical, syntactic) analysis, sampling and error counts, Type 1 suffers from major shortcomings. First, because of time constraints, it cannot assess, except on the basis of statistical probabilities, the acceptability of the content of the translation as discourse. Second, the microtextual analysis inevitably hinders any serious assessment of the content macrostructure of the translation. Third, the establishment of an acceptability threshold based on a specific number of errors is vulnerable to criticism both theoretically and in the marketplace. Type 2 cannot offer a cogent acceptability threshold either, precisely because it does not propose error weighting and quantification for individual texts. Another drawback of Type 2 is that it is applied almost exclusively to literary and advertising texts. What is needed is a model that combines the quantitative and textological dimensions, along the lines proposed by Bensoussan and Rosenhouse (1990) and Larose (1987, 1998), and targets instrumental (pragmatic) translations. The goal of this thesis is to develop an argumentation-centred model to meet this need. With Toulmin's argument macrostructure (1984) as a starting point, a set of argumentation parameters are developed: macrostructure, propositional functions (Widdowson 1978) and reasoning structure (Thomas 1986), conjunctives and inference indicators (Halliday and Hasan 1976; Roulet et al. 1985; and Thomas 1986), argument types (Corbett and Connors 1999), figures of speech, and narrative strategy (Ouellet 1992). These parameters provide a framework for assessing the degree to which translations preserve the argumentation, coherence and cohesion of the message of the source text, for determining critical passages for TQA, and for proposing a new breakdown of errors based on three levels of seriousness. A weighting grid is then added to the model to enable evaluators to take account of conventional parameters such as target-language errors. Finally, a new set of quality standards is proposed for instrumental texts produced by professionals and students alike