Abstract Fluency is an important, yet insufficiently understood, construct in interpreting studies. This article reports on an empirical study which explored the relationship between utterance fluency measures and raters’ perceived fluency ratings of English/Chinese consecutive interpreting. It also examined whether such relationship was consistent across interpreting directions and rater types. The results partially supported the categorization of utterance fluency into breakdown, speed and repair fluency. It was also found that mean length of unfilled pauses, phonation time ratio, mean length of run and speech rate had fairly strong correlations with perceived fluency ratings in both interpreting directions and across rater types. Among a number of competing regression models that were built to predict raters’ fluency ratings, a parsimonious model, using mean length of unfilled pauses and mean length of run as predictors, accounted for about 60% of the variance of fluency ratings in both directions and across rater types. These results are expected to help create, rewrite and modify rubrics and scalar descriptors of fluency scales in rater-mediated interpretation assessment and to inform automated scoring of fluency in interpreting.

中文翻译：

---

连续口译中话语流利度与评分者感知流利度的关系建模

摘要 流利度是口译研究中一个重要但未被充分理解的结构。本文报道了一项实证研究，该研究探讨了英语/汉语交替传译中话语流利度测量与评估者感知流利度评分之间的关​​系。它还检查了这种关系在解释方向和评估者类型之间是否一致。结果部分支持将话语流利度分类为故障流利度、速度流利度和修复流利度。研究还发现，未填充停顿的平均长度、发声时间比率、平均跑步长度和语速与口译方向和跨评分者类型的感知流利度评分具有相当强的相关性。在为预测评分者的流利程度而建立的许多相互竞争的回归模型中，一个简约的模型，使用未填充停顿的平均长度和运行的平均长度作为预测因子，占流利度评分在两个方向和两个方向的方差的约 60%。跨评估者类型。这些结果有望帮助创建、重写和修改评估者介导的口译评估中流利量表的量表和标量描述符，并为口译流利度的自动评分提供信息。占流利度评分在两个方向和跨评分者类型的方差的 60% 左右。这些结果有望帮助创建、重写和修改评估者介导的口译评估中流利量表的量表和标量描述符，并为口译流利度的自动评分提供信息。占流利度评分在两个方向和跨评分者类型的方差的 60% 左右。这些结果有望帮助创建、重写和修改评估者介导的口译评估中流利量表的量表和标量描述符，并为口译流利度的自动评分提供信息。