An experimental use of the ‘default translation’ concept to compare output of professional translators, translation trainees and neural machine translation
| Authors | |
|---|---|
| Year of publication | 2022 |
| Type | Appeared in Conference without Proceedings |
| MU Faculty or unit | |
| Citation | |
| Description | The study is based on an experimental extension/application of the concept of ‘default translation’ proposed by Halverson (2015, 2019) as a part of the non-computational approach to study cognition in translation and interpreting, in order to study and compare output of professional translators, translation trainees and neural machine translation on the English-Czech language interface. In full recognition of Halverson’s proposal of ‘default translation’ as a construct grounded in understanding cognition – including cognition in translation – as embodied, embedded, enacted, extended and affective (4EA) and hence the need for re-focusing attention from formerly prioritized areas of predicted difficulty and conscious deliberation (problem-solving in translation) to translation as an ‘uninterrupted’ and ‘unconstrained’ mode of production, also backed by a considerable body of translation process research, this study adopts an experimental approach to the concept. It attempts to gauge – both theoretically and practically – the possibilities and limitations of ‘transplanting’ default translation beyond the paradigm based on using eye-tracking and keystroke logging methodologies, and identifying default translation phases in translation professionals instead through sight translation tasks. Secondly, translation output and its quality in default translation tasks will be compared for professional translators, translation trainees and several neural machine translation systems (where success at default translation might be considered a base-line level of success). Attention is given to a post-hoc description of linguistic properties of stretches of texts accommodating default translation between English and Czech as less formally similar languages. Degrees of general and syntactic literalness and acceptability of the translational outputs in the three translating agents are compared, with a view to conceptualizing differences in the nature and quality of output of human and machine translation in this base-line difficulty of translation. |
| Related projects: |