***************** Non-literal language use in machine and human translation for dubbing
WRoCAH funded Collaborative Doctoral Award between University of Sheffield, and Zoo Digital Group plc For more information about this project contact Dr Nicole Baumgarten (n.baumgar...@sheffield.ac.uk) Mr Gareth Simpson (gareth.simp...@zoodigital.com) More information at: https://docs.google.com/document/d/1p1D8SDERfc2ra4Yqm5IjnTAe7HrGcdF34dI7Hs8wOzY/edit# ****************** *Project summary The project is a collaboration between the University of Sheffield and Sheffield-based ZOO Digital Group plc, one of the industry leaders in media localisation. The project will investigate non-literal language use (such as idioms, metaphor, jokes, conventional indirectness, or conversational implicatures) in machine and human audiovisual translation for commercial lip-sync dubbing of films. By means of an innovative and original multi-methods approach, the project will close the gap between computational and humanistic approaches to translation and make use of their complementary strengths in the development of an integrated, industry-relevant human-machine translation system for dubbing that is able to handle implicit meaning in film. *PhD Project description Non-literal language features in film are problematic for existing machine translation (MT) approaches, for example because they may have culture-specific implied meanings that MT systems cannot infer, they may be unique ideas originated by the speaker spontaneously (creative language use, linguistic innovation), or their intended meaning may rely on the extended linguistic or visual context in which the utterance is embedded. MT approaches, which rely on linguistic surface structures to identify language patterns, struggle with these kinds of non-compositional, pragmatic and context-dependent meanings, which, however, play a central role in film because they drive forward the story and are crucial to character development. Human dubbing translators, too, face challenges with non-literal language, albeit on the level of re-creating it in equivalent and lip-synchronous ways in the target language, and not on the level of recognising it in the original text. The project will integrate the distinct qualities of MT and the human translator to identify linguistic, textual and visual characteristics of non-literal language in film, and express them as heuristics or algorithms a computer could interpret, so that it might be possible to improve MT, or at the very least identify areas where MT is expected to struggle to direct human translators’ effort. The project taps into the grey area between two distinct views of translation, namely translation as a culture-based interpretive-creative process and translation as a computational-statistical procedure. As a consequence, research on human translation and MT have become increasingly uninformed by one another. The project aims to offer a solution to a specific pervasive problem identified in professional AVT practice by integrating humanistic and MT approaches to translation. It is driven by three broad research questions, which can be shaped by the student candidate, for example with respect to the languages to be considered: 1. How do MT systems and human translators handle non-literal language use in film? 2. How can MT and human translation processes be integrated in AVT? 3. How can MT systems be enhanced by formal descriptions of non-literal language use in film? The project will include multiple methods and analysis dimensions including: ● Analysis of non-literal language in film, with the aim of formalising non-literal language as a film and dubbing text phenomenon. ZOO Digital will provide a set of films, original and dubbing scripts, which allow application of corpus linguistic methods as well as qualitative linguistic analysis. ● Linguistic-pragmatic analysis of translations of non-literal language by MT systems in order to understand the limitations of the current systems. ● An observational study of human dubbing translators’ behaviours as they translate different types of context-embedded non-literal language. This should include self-report methods (for example think aloud protocols, stimulated recall) and direct measures (for example eye-tracking, keylogging, screen recording). The observational study will provide evidence of text- and language-based decision-making procedures and contribute a human processing perspective to the formal description of non-literal language. The overall outcome of the project will be a formal description of types of non-literal language use that can be operationalised for MT and a refined understanding of the role of the human translator in relation to MT and how they can be integrated in a human-machine translation system for dubbing. The project is embedded in an established Knowledge Exchange partnership between researchers at Sheffield and ZOO. *Financial support Studentships for doctoral research are 40 months in duration for full-time study. Awards are subject to satisfactory academic progress. Awards must be taken up in October 2021; no deferrals are possible. Awards will comprise fees at Research Council rates and a maintenance grant (£15,609 in 2021/22). The grant pays the fees at the Home/UK rate; international students are, however, eligible to apply for this Studentship and the difference between the Home/UK and International fee will be met by the University of Sheffield for a successful international applicant. Awards may be taken up on a part-time basis if a student is eligible to undertake part-time study; international applicants may be required to study full-time by the terms of their visa. *Qualifications Strong applicants will have a good first degree in a modern language or linguistics and will be a proficient (near-native or native) speaker of at least one non-English language. Ideally, applicants will have a Masters degree in audiovisual translation or computational linguistics (or be working towards one) and/or relevant industry experience in the field of translation or media localisation. The project will involve the use of machine translation systems, so an aptitude for working with computer systems is required. As the project is interdisciplinary, applicants will have the ability to transcend disciplinary boundaries and bring creative and flexible thinking to their work. *Requirements of the Studentship WRoCAH students are required to undertake a bespoke training package and to complete a Researcher Employability Project of at least a month, an Internationalisation Project and a Knowledge Exchange Project. All WRoCAH students must submit their thesis for examination with the funded period. This is a requirement of the Arts & Humanities Research Council, which provides the funding for WRoCAH, and is a condition of accepting a Studentship. *How to apply By 5pm Friday 9 April 2021, applicants are required to submit to WRoCAH an Expression of Interest, which should include: A CV with details of academic qualifications A covering letter comprising a two-page statement to convey your motivation and enthusiasm for the project, and to demonstrate your suitability for your intended PhD studies with the University and Project Partner. The covering letter should specifically highlight the following: Your interest in the project and details on why you have chosen that University and Project Partner. How you will apply your current skills, knowledge and experience to undertake a PhD and the approach you would take to develop the project. How the project fits into your career plans and ambitions. 16 April 2021: decision on short-listing The short-list of candidates to be invited for an interview will be announced on Friday 16 April 2021. Short-listed candidates must complete a PhD programme application before interview. If you are short-listed for an interview you will be sent details of how to apply for a place at the University of Sheffield. At that point you will need to submit the names and contact details of two referees, copies of transcripts of your academic qualifications and (if applicable) an IELTS certificate. w/c 26 April 2021: interviews Interviews will involve the academic supervisor, Project Partner supervisor and a member of the WRoCAH Studentships Committee. They will be conducted online. For more information about this project contact Dr Nicole Baumgarten (n.baumgar...@sheffield.ac.uk) Mr Gareth Simpson (gareth.simp...@zoodigital.com) -- Carolina Scarton Academic Fellow Department of Computer Science University of Sheffield http://staffwww.dcs.shef.ac.uk/people/C.Scarton/ _______________________________________________ Moses-support mailing list Moses-support@mit.edu http://mailman.mit.edu/mailman/listinfo/moses-support
