Machine Translation in Subtitling: New Job Profiles, New Skills, New Era

Machine Translation in Subtitles

Audiovisual translation (AVT) — particularly in subtitling and dubbing — has received a lot of attention lately, mainly due to the unprecedented growth in digital video consumption and the subsequent need for localization.

Yota Georgakopoulou, Audiovisual Localization Consultant at Athena Consultancy, wrote in a 2020 paper how the world is entering an era of commoditization for certain types of AVT content, “as is already happening in other markets, and as such its production will become more streamlined and standardized in an attempt to make it more efficient.”

This will be facilitated by the integration of language technologies, such as automatic speech recognition (ASR) and machine translation (MT), into traditional localization workflows. As a result, Georgakopoulou predicted, project managers specializing in such workflows and post-editors of ASR and MT will be in high demand in the near future.

For MT to be successfully implemented in subtitling, a number of post-editing skills are required, according to another study.

Moreover, in subtitling, “linguistic competence, sociocultural awareness, and subject knowledge are no longer sufficient,” pointed out Jorge Díaz-Cintas, Professor of Translation Studies at University College London, and Aline Remael, Professor Emerita of Translation Theory and Audiovisual Translation at the University of Antwerp.

Today, along with translation, post-editing, and subtitling competencies, subtitlers need fundamental technical-methodological competencies in MT and ASR. They need to be well-versed in cutting-edge technologies, demonstrate high technical know-how, and be able to quickly adapt to new programs and specifications.

In a 2021 study, Anke Tardel, Research Fellow at the Johannes Gutenberg University of Mainz, Silvia Hansen-Schirra, Professor at the Johannes Gutenberg University of Mainz, and Jean Nitzke, Associate Professor at the University of Agder, identified the skills post-editors in subtitling industry must have as well as three possible job profiles.

The Subtitle Post-Editor

The first job profile is that of subtitle post-editor; that is, subtitlers with special training in post-editing or trained post-editors from the text translation industry with special training in subtitling and AVT.

The skills post-editors working with text translation need were outlined in a recent Slator story. These include basic translation competencies (i.e., bilingual, extralinguistic, and research competence), as well as error-spotting, -classification, and -handling competencies.

Subtitle post-editors also need subtitle-specific competencies, such as spotting, reduction, segmenting, adjusting from speech to written text, and matching of the text with the image. However, some of these competencies may be less relevant when working with template files (i.e., when the spotting and segmentation of subtitles is already set) and post-editing is performed within the constraints of the template. I Recruit Talent. Find Jobs

LocJobs is the new language industry talent hub, where candidates connect to new opportunities and employers find the most qualified professionals in the translation and localization industry. I Recruit Talent. Find Jobs

Basic MT knowledge might also be helpful for subtitle post-editors. Knowing how MT systems work, which possible pitfalls each system may generate, and which system was deployed in each case can help them spot potential errors more easily.

When MT is used in combination with ASR, recognition errors from the ASR might be transferred as well. Thus, subtitle post-editors should also be familiar with the applied ASR technology, know which errors to expect — such as adjusting errors in automatic timing, compression, and segmentation — and how to correct them most efficiently.

The ASR & MT Engineer for Subtitling

ASR & MT engineers for subtitling need specialist competencies. In particular, they should have a deep knowledge of ASR and MT systems architecture and requirements, good programming skills, as well as in-depth knowledge on how to train and evaluate those systems so that they can be appropriately tailored to the subtitle post-editor’s needs.

To do so, they should have a thorough understanding of the subtitles’ nature — synchronized, condensed, and segmented text across lines and subtitles — as well as the differences between speech and written text in general. In addition, they should be familiar with different style guides, which may vary depending on the target language, medium, broadcaster, or streaming provider. 

Finally, ASR & MT engineers should be aware of the various subtitle file formats, as well as the availability and quality of training data (i.e., aligned subtitle files).

The MT Consultant for Subtitling

A consultant’s main job involves assessing project requirements, analyzing potential risks, and making strategic decisions. In order to perform the necessary risk management and proper consulting, MT consultants for subtitling should know the AVT market very well, have a good understanding of the subtitled media content — including knowledge on film rights, genres, and processes of film production and distribution — as well as basic knowledge of ASR and MT engineering.

Once again, basic translation and post-editing competencies, and soft skills are required to fully comprehend the entire process and to effectively consult language service providers and film producers or distributors alike on when and how to apply machine translation and post-editing.

According to the authors of the aforementioned study, this role would be ideal for project managers working in AVT with training on post-editing.

It is obvious that technology has brought changes to the AVT industry and affected the way professionals work. These changes may prove unsettling for some, while others may opt not to keep pace — “but those that are responsive and alert to the future directions of the AVT industry will find themselves part of an exciting journey,” according to Lindsay Bywood, Senior Lecturer in Translation Studies at University of Westminster.