One of the main research questions was how translation performance changed during training. Because of data loss and non-compliance (see above), 10 human translators’ and 13 post-editors’ development could be analysed. Table 29 gives details about how human translators’ performance changed between the first and second year of training. Closer inspection of the table shows that, in addition to a drop in the total number of errors, there was a decrease in accuracy and fluency errors. However, the number of style errors has not changed dramatically, and terminology errors even show a slight increase between the beginning and the end of the training. Nevertheless, none of these changes are statistically significant, although accuracy errors are trending towards significance.

 

 

Differences between post-editors’ first- and second-year performance indices are presented in Table 30. It is apparent that patterns of changes are very similar to those in the HT group: accuracy and fluency errors decreased in number, together with the total number of errors. Style errors dropped slightly, and terminology errors increased. However, in the PE group, changes in fluency and terminology errors, and in the total number of errors were significant. Whereas the decrease in the total number of errors and in any other error types does not come as a surprise, an increase in terminological errors is an unexpected finding. Possible interpretations of this result will be offered in 3.6.2.2 Differences in error numbers between the first and the second data collection waves.

 

 

In addition, we wanted to see whether the magnitude of the changes was significantly different in the two groups (i.e., whether either of the groups improved more on any of the factors). To achieve this, variables were created by subtracting the number of second-wave errors from the number of first-wave errors (e.g., ‘Accuracy_change’ = Acc1–Acc2), and independent samples t-test were performed on these variables. The tests indicated no significant differences between the HT and PE group, which means that the improvement of the groups was comparable (Table 31).