ABSTRACT
Quality assessment of human translations includes a revision step, in which erroneous text spans in target documents are identified as translation issues with particular revision proposals. The identified issues are also classified into abstract issue types, such as “omission” and “misspelling,” for in-depth evaluation of translations and translators. This chapter presents our method in designing an issue classification scheme. Aiming at (i) a consistent human assessment, (ii) of English-to-Japanese translations, (iii) produced by learner translators, we composed our issue classification scheme in the form of a list of issue types, i.e. a metalanguage of translation issues, accompanied by a decision tree for consistent use of the metalanguage, and manually tuned them through the application of the OntoNotes method, i.e. an iteration of assessing learners’ translations and hypothesising the conditions for consistent decision-making, as well as refining the metalanguage. Intrinsic evaluation of the created scheme confirmed its potential contribution to the consistent classification of identified issues, leading to a substantial inter-annotator agreement. This chapter also describes potential applications of issue classification schemes in general and future research directions to make them more useful and practical, thereby facilitating communication between translators (learners) and revisers (instructors).
