Improving Products and Processes in Translation Technology Use (IMPETUS)

Lead Research Organisation: University of Bristol
Department Name: School of Modern Languages

Abstract

Recent annual figures show that language services are a US$46-billion industry with high growth forecasts. Technology is crucial for coping with demand in this sector. In many contexts, translators now edit and interact with machine translation (i.e. systems akin to Google Translate) instead of translating texts from scratch. The productivity benefits of using machine translation in this way are well known, but the process is controversial and far from smooth. Attitudes to machine translation among language professionals are famously negative. In the press, translators are said to 'have the blues' (The Economist 2017) and to be 'concerned' (El Pais 2017) about the impact of machine translation on their jobs and work. Methods for training translators to use technology are often ineffective, and knowledge of how machine translation affects translators' linguistic decisions is still limited, which poses obstacles to a beneficial use of machine translation in professional workflows. This project is a joint effort to address these issues and improve products and processes linked to translators' interaction with technology. We will examine the efficacy of different working methods and assess new training approaches that can be deployed in the translation workplace. In a context of heated debates on the impact of automation on human labour, our aim is ultimately to empower translators to adopt machine translation in more effective and rewarding ways that can improve their job satisfaction, raise professional standards and boost the multilingual economy.

The project will have three phases. In the first phase, we will investigate the use of neural machine translation, a new paradigm in machine translation technology, and assess different strategies that translators can adopt to improve their work. The second phase will delve deeper into links between translating behaviour and the quality of translated texts. We will examine these links across languages in the largest-scale study to date to establish a connection between what translators do and the quality of the texts they produce. In an innovative integration of research methods and business practices, the third phase of the project will examine how tools that track translators' activity (e.g. edits and translating speed) can be used positively and responsibly in the workplace. We will monitor translators' activity for four months to shed new light on issues such as data ethics and perceptions of productivity as well as on the practical implications of incorporating methods from the project's initial phases into commercial contexts.

Throughout the project, we will use a unique combination of methodologies including objective and subjective techniques. The work is novel in several ways. First, it will provide best-practice documentation on the use of activity tracking in commercial settings and on key editing procedures so far ignored by industry standards and guidelines. Second, with a diverse team of researchers and industry partners, the project constitutes a rare collaboration between translators, academics, technology developers, translation businesses and professional bodies. Our work is rooted in problems reported by and discussed directly with our professional partners. Finally, we will veer away from previous initiatives focused predominantly on the development of new tools to represent a much-needed investment in know-how and capacity building. To date, translators' concerns about the profession have not necessarily aligned with top-down investments in technology that often disregard the technology users, with potentially negative consequences for the technology developers and for translators' professional standing. We hope to change this by concentrating on the human aspects of technological progress and providing empirical evidence on new ways of working and the efficacy of different professional practices.

Planned Impact

Who will benefit from the research?

1. Professional translators

2. Translation company owners and managers

3. Developers of translation technology and of translation technology training

4. Translation students and trainers

How will they benefit from the research?

1. The Institute of Translation and Interpreting (ITI), a globally renowned professional association for translators, is one of the project's partners (see letter of support). The project findings will be shared with ITI's Research Network and disseminated among ITI members through co-produced outputs and events. This will influence translators' approach to technology. They will be able to adopt new ways of working based on these discussions, for example by using tracking tools to gain awareness of their processes. The process of achieving these changes is already in motion through interviews carried out in preparation for the project (see Case for Support). A report on the interviews is available on ITI's website, where the University of Bristol has a dedicated space (https://www.iti.org.uk/professional-development/research/university-of-bristol). The project will involve independent professionals and address issues of major societal concern. An anonymous translator who attended a sold-out knowledge exchange event to discuss the interview findings mentioned on the exit form that the event was 'comforting for his/her MT [machine translation] fears'. Another one stated that it provided 'a possible way forward'. The event was rated 4/4 by most of those who completed the form.

2. Following new and updated information on best professional practices, translation companies will be able to improve their business strategy. The preliminary interviews indicated a misalignment between many complaints of the participating companies and some of their most frequent practices. Translation production issues mentioned by managers often concerned how translators approached translation tasks (e.g. where they over-edited the text). However, when asked about how they were attempting to tackle these problems, the companies often reported standard error-counting techniques that do nothing to improve translators' approach. The project will allow process-related problems to be addressed based on empirical evidence. Tatutrad (Partner 2 - a translation company in Spain) will be on the front-line of some of these initiatives. To ensure a wider application of the findings, we will be sponsors of a major industry event and exploit our sponsor status to invite other businesses to test our recommendations and report to us on their impact.

3. Technology developers will be able to enhance their products based on the project findings. SDL Plc (Partner 1 - a major provider of language services, technology and training) will benefit from the project first-hand by being able to: (a) update online machine translation post-editing certification tests; (b) fine-tune their tools and machine translation systems based on empirical results reflecting translators' behaviour; and (c) provide their clients with evidence-based information on products and practices.

4. Students and trainers of translation, including those affiliated to the over 550 institutions who participate in SDL's academic programme, will benefit from the research by making use of training materials reflecting empirically sound information on translating behaviour and state-of-the-art technology. The PI has contributed to a 'training the trainers' e-book that SDL currently distributes to universities. The project findings will feed into this documentation and into teaching practices. Just within the project team, these initiatives will reach three countries and institutions (University of Bristol, Kent State University and Universidad Pablo de Olavide). They will be disseminated further, for example through SDL's academic programme and through our participation in major events attended by the academic community.

Publications

10 25 50