ABSTRACT

Investigating the coherence of translated texts is an important issue in multilingual studies. In this paper, we aim to study text coherence in human translated texts and its relation to the text properties by a quantitative approach. For this purpose, we assigned a word importance value to each word-type of a text and construct the text ‘importance time series’ from the original and translated texts. Then, we calculated text global coherence by applying Detrended Fluctuation Analysis (DFA) to these time series. By means of this procedure, we were able to compare the coherence of the original and translated texts. Our results show that a translation does not always decrease text coherence, as many people may suppose; there are many cases where text coherence is increased by translation. We also studied the relation of text coherence and the text properties such as text size or vocabulary size; we observed no relevance. Our findings suggest that the coherence of a text depends on the translator’s abilities rather than the state of being original or translated.

摘要

调查翻译文本的连贯性是多语言研究中的一个重要问题。在本文中，我们旨在通过定量方法研究人工翻译文本中的文本连贯性及其与文本属性的关系。为此，我们为文本的每个词类型分配了一个词的重要性值，并从原文和翻译文本中构建了文本“重要性时间序列”。然后，我们通过对这些时间序列应用去趋势波动分析 (DFA) 来计算文本全局连贯性。通过这个程序，我们能够比较原文和翻译文本的连贯性。我们的结果表明，翻译并不总是会降低文本的连贯性，正如许多人所想的那样；在很多情况下，翻译会增加文本的连贯性。我们还研究了文本连贯性与文本属性（如文本大小或词汇大小）的关系；我们观察到没有相关性。我们的研究结果表明，文本的连贯性取决于译者的能力，而不是原创或翻译的状态。