Empowering public service interpreters with the help of automatic speech recognition

This project will maximise the impact of recently completed InnovateUK funded research ('Harnessing court data using NLP and spoken language technology' REF:10022430) by embedding its research outcomes in the development of bespoke Automatic Speech Recognition (ASR) services for interpreters and providing training in using ASR. The focus will be on supporting public service interpreters (e.g. interpreters working in healthcare, legal, local government settings), who, in contrast to conference interpreters, currently benefit much less from the latest developments in artificial intelligence (AI).

Once our solution is market-ready, we will offer bespoke ASR engines for interpreters, as well as training in customising and using ASR engines. We will also explore how to extend the project to language service providers (LSPs) and users of interpreting services (UIS) in the future.

As a result of recent developments in AI, the field of ASR has advanced rapidly, leading to more accurate rendering of speech for general domains and easier access to ASR for everyday users. Research conducted at the Centre for Translation Studies (CTS) showed that interpreters, including public service interpreters, could benefit from integrating ASR in their work.

However, many interpreters do not have the expertise to use ASR, to access ASR services and to customise them to obtain better accuracy in well-defined domains.

In our previous research, we showed how Natural Language Processing (NLP) techniques can be used to improve the accuracy of ASR engines for the legal domain. The goal of this project is to validate and deliver a sustainable solution for developing and accessing customised ASR engines for interpreters. To achieve this, we will streamline the process of ASR customisation for a specific domain and language by preparing a suite of easy-to-use programs for collecting domain-specific data, customising and using ASR engines like Amazon Transcribe. The tools will then be offered to interpreters.

In addition, the project will develop a training course focusing on ASR literacy for interpreters. It will educate them on how ASR works, how interpreters can benefit from this technology and when it is appropriate to use it. The course will also include a practical part where participants will learn how to use our services to customise an ASR engine.

In parallel, the project will identify the most appropriate business model to build a sustainable service that meets the needs of individual interpreters and, in the future, LSPs and UIS, ensuring commercial activity in the medium and long term.

The project team delivering this work comprises researchers, commercial projects officers, and educators, and will be supported by a wider academic environment at the University of Surrey that is ideal for research commercialisation. Just:access, the company that was involved in the original project will provide advice regarding the potential commercialisation avenues.

The Institute for Translation and Interpreting (ITI) and the Chartered Institute of Linguists (CIoL) will support this project by helping to reach out to interpreters, LSPs and UIS, enabling us to better understand the market landscape. They will also be involved in shaping of the training materials.