机器翻译质量综合评价方法研究

Study on the Comprehensive Evaluation Method of Machine Translation Quality

基于对"词汇"、"语法"和"语篇"三个译文质量评价因素的分析,本文构建了包含12个指标、每个指标分为5个质量等级的评价指标体系,通过问卷设计与问卷调查方法获取机器翻译质量的评价数据。采用层次分析法确定评价指标的权重向量,利用模糊数学理论建立了机器翻译质量的综合评价向量模型。实例分析表明,该模型可以科学定量地评价翻译软件的译文质量及其在各项评价指标上的差异,对读者筛选优质翻译软件、开发商提高软件质量具有重要意义。

Based on an analysis of three factors in the assessment of translation quality, i.e. vocabulary, gram- mar and discourse, this paper developes an evaluation index system which includes twelve indicators and each indicator is divided into five levels. Meanwhile, a questionnaire is also designed for use. Foreign language teachers and Master＇s degree candidates are selected as respondents. Then, the data for machine translation e- valuation is obtained from the questionnaire. The analytic hierarchy process （AHP） is applied to the data in or- der to determine the weighted vector of machine translation evaluation factors. On this basis, fuzzy mathematics theory is used to establish the comprehensive evaluation vector for machine translation quality. As a result, the quantity comprehensive evaluation for the translation text quality is completed. The results of our study show that the comprehensive evaluation model of translation quality based on the AHP and fuzzy mathematics method can be employed to scientifically and quantitatively analyze the translation text quality of the translation software as well as the differences of various evaluation indicators. Thus, this is significant for readers to choose high quality translation software and for software developers to find design flaws so as to improve the software design and the quality of software.