The results presented here have been produced using a sub-set of the total number of sentences and TIs under analysis. Nonetheless, some tentative findings can be highlighted and it is expected that analysis of the remaining sentences and TIs will further underpin the findings presented here. The data suggest that the post-editing speed for sentences with minimal TIs is, on average, faster than for sentences with TIs. This implies that the application of CL rules to improve machine translatability will indeed reduce post-editing effort. However, the data also suggest that some TIs, e.g. long NPs and the gerund, cause post-editing effort while others, e.g. abbreviations and proper nouns, do not. Further analysis will allow us to specify exactly which TIs are most problematic. It is hoped that this kind of information could be used in the tuning of CL rules in the future. Finally, although a sentence may not contain a TI, the data suggest that such a sentence may still be subject to post-editing activity. One of the aims of the ongoing research is to identify the reason(s) why such sentences are being post-edited.
展开▼
