Assessing Quality Of Machine-generated Bilingual Subtitles

In the digital era,videos,as an important channel of information,have been circulating the globe,driving up the need for subtitling and translation.In the face of the surge in market demand,the subtitling and translation industry should innovate by using high technologies.Currently,translation technologies,including ComputerAided Translation and Machine Translation,have been primarily confined to the traditional localisation industry,instead of the field of subtitling,especially in China.Thus,this study aims to explore an emerging Chinese automatic subtitling platform:“Net Ease Sight”,by evaluating the quality of its machine-generated bilingual subtitles through the Word Error Rate and FAR model,accompanied by discussions on its advantages and disadvantages.It was found that the English subtitles are highly accurate,but the sentence segmentation was often inappropriate,leading to winded or broken sentences.In addition,the overall quality of the Chinese subtitles is low,with a number of semantic errors and a few grammatical and idiomatic mistakes caused by poor lexical choice,rigid verbatim translation,lack of context and audience awareness.More importantly,the overall low quality is essentially due to poor sentence segmentation and misrecognition of the original speech.Therefore,in order to ensure the quality of Chinese and English output,the technologies of Automatic Speech Recognition,Automatic Timecode Generation and Machine Translation should be enhanced hand-in-hand.Besides,there are also promising signs regarding machine-generated subtitles,including good sound-text synchronisation,correct language and punctuation use,and natural language style.Therefore,MachineAssisted Subtitling can be a future trend,where machine-generated subtitles and human post-editing are combined,which will reduce the workload and cost of human labour while guaranteeing the efficiency and quality of subtitles.