Machine Translation (MT) technology has been widely used in the localisation industry to boost the productivity of professional translators. However, due to the high quality of translation expected, the translation performance of an MT system in isolation is less than satisfactory due to various generated errors. This study focuses on translation of prepositions from English into Chinese within technical documents in an industrial localisation context. The aim of the study is to reveal the salient errors in the translation of prepositions and to explore possible methods to remedy these errors.udThis study proposes three new approaches to improve the translation of prepositions. All approaches attempt to make use of the strengths of the two most popular MT architectures at the moment: Rule-Based MT (RBMT) and Statistical MT (SMT). The approaches include: firstly building an automatic preposition dictionary for the RBMT system; secondly exploring and modifing the process of Statistical Post-Editing (SPE) and thirdly pre-processing the source texts to better suit the RBMT system. Overall evaluation results (both human evaluation and automatic evaluation) show the potential of our new approaches in improving the translation of prepositions. In addition, the current study also reveals a new function of automatic metrics in assisting researchers to obtain more valid or purpose-specific human valuation results.ud
展开▼
