基于语料库的人机文学译本语言特征对比研究——以《傲慢与偏见》三个译本为例

A Corpus-based Comparison of General Language Features of Human Translation and Online Machine Translation

本研究基于由《傲慢与偏见》两个人工译本与一个在线机译本及其英语原文构成的平行语料库,从词法和句法层面对比了人机译本的16个语言计量特征及风格特征。结果表明,在线译本在词法、句法和情感表达等方面都远不如人工译本,在代词、助词和连词使用上"显化"明显,在标准类型比和词汇密度方面"简化"不足,而句法层面无"明晰化"倾向。这与其在词性转换、上下文语境判断、情感识别与时态标记识别等方面能力不足有关。两个人工译本在词汇和句法层面也呈现明显的风格差异,其"显化"、"简化"和句法"明晰化"的程度也各有不同,其成因与不同的翻译目的和译者主体性有关。

Based on a parallel-corpus of Pride and Prejudice consisting of two human translations, one online-machine translation and the original text, this study compares the 16 quantitative and stylistic features of the human translations with those of the online translation. Results of the study indicate that the online translation is far behind human translations at lexicon and syntactic levels and in sentiment expression. It is abundant in features of ＂explicitation＂ such as using pronouns, auxiliaries and connections but it lacks ＂simplicity＂ in standard type/token ratio and lexical density, and fails to show the tendency of syntactic ＂explicitation＂. These may be attributed to its weakness in identification and transformation of parts of speech and identification of the contextual meanings, semantic sentiments and tense of the original sentences. The two human translations show different lexical and syntactic features and differ in degree of ＂explicitation＂, ＂simplification＂ and syntactic ＂explicitation＂, which are probably due to the different translation purposes and subjectivities of the translators.