论文部分内容阅读
N元组匹配和翻译单位对齐各有优劣。本文采用320篇学生英译汉译文探讨了它们在自动评分中的取舍问题。研究比较了N元组匹配数量和翻译单位对齐数量与人工对译文语义、形式、总评分的相关性,并采用多元回归考察了它们对译文质量的解释力。结果表明:(i)翻译单位对齐数量与人工评分的相关度高于绝大多数词-和字-N元组匹配数量;(ii)与N元组匹配数量的整体作用相比,翻译单位对齐数量对语义评分的解释力稍高,对形式和总评分的解释力稍低;(iii)与仅以N元组匹配数量或翻译单位对齐数量为自变量的模型相比,词-一元组和翻译单位对齐数量结合产生的模型对人工评分的解释力更强,模型评分与人工评分的相关度和一致性也更高。这表明词-一元组匹配与翻译单位对齐互为补充,两者结合对译文质量的预测效果最佳。
N-tuple matching and translation unit alignment have their own advantages and disadvantages. This article uses 320 English-Chinese translation of students to explore their choice in the automatic scoring problems. The study compared the number of N-tuples matched and the number of translation units aligned with the semantic, formal, and overall ratings of translations, and examined their explanatory power of translation quality using multiple regression. The results show that: (i) the correlation between translation unit alignment and artificial score is higher than that of the vast majority of word-sum-N groups; (ii) compared to the overall effect of the N-tuple matching number, translation unit alignment The explanatory power of the semantic score is slightly higher, while the explanatory power of the formal score and the total score is slightly lower. (Iii) Compared with the model which only matches the number of N-tuples or the number of aligned units of translation unit, The number of translation units aligned with the resulting model is more explanatory to the manual score and the correlation and consistency of the model score with the manual score are also higher. This indicates that word-one-group matching and translation unit alignment complement each other, and the combination of the two makes the best prediction of translation quality.