Error Classification of Machine Translation A Corpus-based Study on Chinese-English Patent Translation

Jiuan-An Hsu

《翻譯學研究集刊》 18輯 (2014/12) Pp. 121-136

並列摘要

While machine translation (MT) systems have been widely applied to translation tasks, the quality of the text outputs often remains unsatisfactory. The demand for better output quality prompts researchers to focus on finding effective ways to evaluate the quality of MT outputs. One popular approach is human error analysis, which is the manual identification and classification of errors made by MT systems. Although there have been many studies examining common error types in MT, none has been found to be targeting the distant language pair of Chinese and English. This study looks into errors in Chinese-English MT of patent abstracts, as such a distant language pair may result in very different error types. In the first level of the hierarchical classification scheme used in this study, errors are split into five major categories: orthographic, morphological, lexical, semantic, and syntactic errors. Each main category is further divided into several subcategories. Thirty-four MT outputs were manually corrected and annotated to identify the distribution of translation errors. The findings suggest that certain features of the Chinese language, such as low occurrences of articles and relatively unclear sentence and phrase structures, do severely affect the performance of the MT system studied. These findings have implications for MT system developers and post-editors.

並列關鍵字

Machine translation ； manual error analysis of MT output ； error classification of Chinese-English MT output

參考文獻

Alam, Y. S.(2013).Manual Evaluation and Error Analysis of Machine Translation Output between a Distant Language Pair Focusing on Effects of Sentence Length.Tenth Symposium on Natural Language Processing (SNLP-2013).(Tenth Symposium on Natural Language Processing (SNLP-2013)).:

Arnold, D.,Balkan, L.,Humphreys, R. L.,Meijer, S.,Sadler, L.(1994).Machine translation: an introductory guide.London:NCC Blackwell.

Banerjee, S.,Lavie, A.(2005).METEOR: An automatic metric for MT evaluation with improved correlation with human judgments.Paper presented at the Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization.(Paper presented at the Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization).

(1994).Reversible Grammar in Natural Language Processing.Springer.

Bennett, P. A.(1981).The evolution of passive and disposal sentences.Journal of Chinese linguistics.9(1),61-90.