The Linguist

@CIOL_Linguists WINTER 2023 The Linguist 11 FEATURES For example, a study conducted with an early prototype boothmate found that the availability of ASR support increases the share of complete renditions (from 67.7% to 90.2%) and drastically reduces the number of omissions (from 15.8% to 3.5%). 1 All A>B experiments conducted to date show an improvement in interpreting quality when such tools are used, revealing that it is possible to integrate additional information into the interpreting workflow. An AI boothmate can be used while interpreters work alone (in the case of short assignments) or with a human boothmate, leveraging what AI can do best: retrieving information. The user experience To enhance the practical application of this technology, studying the optimal user interface (UI/UX) is crucial. In response to this need, the University of Ghent launched a research initiative, Ergonomics for the Artificial Booth Mate (EABM), in partnership with the Johannes Gutenberg University of Mainz, funded by the European Commission. The project conducted a survey using mock-ups to determine the most effective UI/UX designs, specifically focusing on the placement of information on the screen. With 525 interpreters participating – a notable sample size given the specialised field of simultaneous interpretation – the survey highlighted the community's strong interest in harnessing AI to enhance their work. The results favoured a vertical layout where new items appear below previous ones. Items should stay visible for an extended period, with terms positioned on the left and numbers on the right, or both consolidated in a single area. An empirical experiment involving professional interpreters from the European institutions followed to test the design in a real-world setting. The outcomes further validated previous research, suggesting that a meticulously designed artificial boothmate can enhance precision and elevate translation quality. There is now substantial scientific literature supporting the potential of AI to enhance the performance of professional interpreters in certain aspects of their work. However, significant challenges and questions remain. Foremost among these is data confidentiality. Given that many cutting-edge AI tools operate on the cloud, privacy concerns arise – an issue particularly pertinent for professional interpreters. Furthermore, the limited availability of these tools for low-resource languages leads to an imbalance. This means that while some interpreters benefit from AI advancements, others are at a disadvantage. The community should address these challenges, even though it may demand considerable time and effort. Universities, for example, could create projects to collect data in low-resource languages and improve foundation technologies. Because of the importance of such endeavours, funding should be possible. The niche nature of conference interpreting, where this technology first took root, has impeded its widespread development and adoption at scale. Yet the same technology holds promise in other interpreting contexts. In settings such as liaison or consecutive interpreting, and even in non-professional capacities, AI could play a transformative role. Such potential could reignite interest in its study and practical integration. Notes 1 Defrancq, B and Fantinuoli, C (2020) 'Automatic Speech Recognition in the Booth: Assessment of system performance, interpreters' performances and interactions in the context of numbers'. In Target © PEXELS

• Cover