Quality Assessment of Interlingual YouTube Auto-generated Closed Captions in Some Crime Narratives Applying the NTR Model | ||||
| Textual Turnings: An International Peer-Reviewed Journal in English Studies | ||||
| Volume 5, Issue 2, 2023, Page 59-83 PDF (967.08 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/ttaip.2023.340217 | ||||
 View on SCiNiTO | ||||
| Author | ||||
Rania Abdel baky Ali Allam 
| ||||
| Faculty of Languages, October University for Modern Sciences and Arts (MSA), Egypt. | ||||
| Abstract | ||||
| Audiovisual Translation (AVT) is a prolific milieu of cross-cultural communication. YouTube is one of the most prominent user-generated, social media streaming platforms. YouTube employs artificial intelligence (AI) devices namely, automatic speech recognition (ASR) and neural machine translation (NMT) to add auto-generated closed captions (CCs) as interlingual subtitles to its broadcasted videos. The present study attempts to assess the translation quality of these CCs to gauge their reliability as mediating tools enhancing culture and entertainment. Translated, auto-generated CCs on three YouTube videos on true crime channel entitled Twisted Minds are scrutinized in March 2023 applying Romero-Fresco and Pöchhacker's (2017) NTR model. Results show that the translated CCs are accurate with only (95%) approximately with a rate of less than the minimum starting point according to (0/10) scale suggested by the NTR mode. Errors of translation content and form as well as speech recognition errors are spotted, indicating a suboptimal translation quality. The auto-generated CCs display reasonable acceptability in what concerns AVT norms, yet with some deviations. Despite such acceptability and instances of positive effective editions of translational manipulation displayed in the CCs, the profuse errors mar the denotative and connotative meanings of the overall content of the crime narratives exhibiting semantic and pragmatic failures. A revisit analysis for the same data is conducted in December 2023, showing an accuracy rate of (97.31%) approximately with a rate of (3+/10) on the NTR model accuracy scale. Improvement is rather notable, yet the accuracy rate is still poor. This proves that seamless ongoing human intervention on the linguistic, semiotic and technical levels in the performance of the YouTube AI devices is much needed to achieve notable advancements in the quality of its auto-generated translated CCs, to be considered a reliable tool that can help demolish communication barriers. | ||||
| Keywords | ||||
| Audiovisual translation; YouTube closed captions; Artificial Intelligence; Automatic speech recognition; neural machine translation; NTR model | ||||
|
Statistics Article View: 471 PDF Download: 584 |
||||
