Efeitos cognitivos e esforço de processamento de metáforas em tarefas de pós-edição e de tradução humana: uma investigação processual à luz da teoria da relevância

Abstract

Drawing on Relevance Theory (SPERBER; WILSON, 1986/1995), this study is an empirical investigation of the cognitive effort required to post-edit machine translated metaphors and to translate this trope manually. More specifically, we aimed at analysing the impact of context and two different machine translation (MT) outputs on cognitive effort required to post-edit machine-translated metaphors. We also aimed atinvestigating cognitive effort required to post-edit machine-translated metaphors compared to their translation from scratch, analysing eye movement on both metaphors and non-metaphors as well as correlating cognitive effort and cognitive effects. We hypothesized that a) context would determine the amount of effort required to post-editmachine-translated metaphors and to translate them from scratch; b) the statistical MT output would have a positive effect on reducing cognitive effort; c) translating metaphors from scratch would require more cognitive effort than post-editing them; d) post-editing metaphors on Task 2 would need less effort compared to Task 1; and, e) in line with Relevance Theory premises, there would be a trade-off between cognitiveeffects and cognitive effort required to post-edit metaphors and to translate them from scratch. The experimental design consisted of three groups: one control group and two experimental groups. Participants from both experimental groups were asked to postedit a machine-translated journalistic text whereas the control group was assigned to translate it from scratch. Each experimental group had two post-editing tasks. On Task 1 (T1), participants were asked to post-edit a Google machine-translated output whereas on Task 1 (T1) the same participants were assigned to post-edit a Systran machinetranslatedoutput. Data collection was conducted under the experimental paradigm of data triangulation in translation process research (ALVES, 2003). Five research instruments were used, namely a prospective questionnaire to profile the participants, retrospective verbal protocols (free and guided), a five-point Likert Scale, Translog to log keystrokes and mouse movements, and two eye trackers (Tobii T60 and TobiiTX300) to track eye movements on areas of interest. Data analysis relies on eye tracking data (fixation duration and pupil dilation), Likert Scale responses, Translog data (analysis of translation units) as well as guided and free verbalizations. Consistent with Gibbs & Tendahl (2006) and Gibbs (2010) theoretical assumption on metaphorcomprehension, our data analysis confirmed that context (and not the type of metaphors) determines the amount of effort required to both translate and post-edit metaphors. Contrary to our hypothesis, results suggest that the rule-based machine translation output might have a positive impact on reducing cognitive effort required topost-edit machine-translated metaphors. In addition, the analysis of data collected in the experimental groups shows that cognitive effort required to post-edit metaphors is not lower in T2 compared to T1. This result is consistent with the relevance-theoretic inferential paradigm and suggests that the raw machine translation output may have stimulated new inferences, which consequently increased the cognitive effort requiredto post-edit metaphors. Furthermore, contrary to what Krings (1994/2001), OBrien (2006a, 2006c) and Carl et al (2011) have found when analysing the complete task, our analysis found that cognitive effort required to post-edit metaphors is higher than translating them from scratch. These results were unexpected and should be interpretedwith caution since participants from the control group (manual translation) were professionals whereas the experimental groups (post-editing) consisted of students. In terms of trade-off between cognitive effort and cognitive effects, we found three possibilities of interaction: 1) more cognitive effort results in more cognitive effects, 2)more cognitive effort does not result in additional cognitive effects, and 3) less cognitive effort results in more cognitive effects. These results confirm empirically three out of four possibilities of interaction raised theoretically by Gibbs and Tendahl (2006) regarding metaphor comprehension.