Rule-based Automatic Post-processing of SMT Output to Reduce Human Post-editing Effort

摘要

To enhance sharing of knowledge across the language barrier, the ACCEPT project focusesrnon improving machine translation of user-generated content by investigating pre- and posteditingrnstrategies. Within this context, we have developed automatic monolingual post-editingrnrules for French, aimed at correcting frequent errors automatically. The rules were developedrnusing the Acrolinx~(IQ) technology, which relies on shallow linguistic analysis. In this paper, wernpresent an evaluation of these rules, considering their impact on the readability of MT outputrnand their usefulness for subsequent manual post-editing. Results show that the readability of arnhigh proportion of the data is indeed improved when automatic post-editing rules are applied.rnTheir usefulness is confirmed by the fact that a large share of the edits brought about by thernrules are in fact kept by human post-editors. Moreover, results reveal that edits which improvernreadability are not necessarily the same as those preserved by post-editors in the final output,rnhence the importance of considering both readability and post-editing effort in the evaluation ofrnpost-editing strategies.