An Experiment Report On Interpreting Quality Assessment Of IFLYTEK Translator 2.0

According to 2018 China Language Service Industry Development Report,translation technology has entered a stage of prosperity since 2011.As a new product of translation technology and a new interpreting service provider,interpreting device,integrating Neural Machine Translation(NMT),speech recognition and synthesis,has been favored by more and more customers.However,in China,there are very few studies on the interpreting quality and assessing methods or criteria of this kind of interpreting devices.Therefore,this experiment selects iFLYTEK Translator 2.0,a representative product of the industry,as the test subject,making a tentative attempt to assess its interpreting quality,analyze its merits and demerits,and discuss the assessing criteria by quantitative and qualitative analysis.Firstly,studies on interpreting quality assessment both at home and abroad are reviewed,especially those on assessing criteria of human interpreters' interpreting quality.The author sums up three comparatively commonly-used parameters from the previous empirical studies,namely,Fidelity,Delivery,and Language,which are used as assessing items in this experiment.In terms of importance,different weights to the parameters are set respectively as 0.5,0.3,0.2,and five-grade rating scales for each assessing item are also properly designed.Then,tourism and business talks in English and Chinese are selected as test materials for interpreting.Referring to the real-time recordings of iFLYTEK Translator 2.0's interpreting,15 raters are invited to score on Fidelity,Delivery,Language and Overall Quality of the interpreting performance based on the rating scales.Finally,through the statistical results of rating,quantitative and qualitative analysis,and post-interview results with the raters,the thesis draws the following conclusions: the overall quality of iFLYTEK Translator 2.0's interpreting is widely recognized by the raters and slightly beyond their expectations;iFLYTEK Translator 2.0 performs best on Fidelity,and has great advantages in responding speed and terminology interpreting;it also has the ability to make some self-correction.However,there are still deficiencies in interpreting culture-loaded words and expressions in specific contexts;speech recognition failures and pauses exert a negative impact on Fidelity and Delivery;there are also problems with pronunciation and intonation.And from the result of t-test and rater's feedback,the assessing criterion has a certain degree of rationality.The preliminary attempts made in this experiment fill the gap in domestic-related researches to some extent;the assessing criteria and methods in the research could also offer references for further study.Additionally,referring to the merits and demerits of Translator 2.0,language services providers and consumers can have a more comprehensive and rational understanding of such products.The findings also provide a direction for further development and improvement of interpreting devices.