This is a pilot course organised part of the Empowering public service interpreters with the help of automatic speech recognition project that focused on how AI and ASR in particular can benefit interpreters, language service providers and companies. It was followed by interviews with course participants in order to collect their opinions about the topic covered and understand their training needs. The aim of the course was to inform the development of a larger CPD programme on “Interpreting and AI”
Session 1: Introduction to AI and how interpreters can benefit from AI (Wednesday, 22nd Jan @ 5pm GMT)
The aim of the first session is to understand the fundamentals of AI and its relevance to interpreting. The session will explain terms like Artificial Intelligence (AI), Generative AI (GenAI), Large Language Models (LLMs), Automatic Speech Recognition (ASR). It will also provide non-technical explanation of how LLMs work and give participants the opportunity to practice with Google’s Gemini. We will use Gemini in this course because it powers NotebookLM which will be used in the 3rd session. However, the knowledge acquired will be applicable to other LLMs like ChatGPT as well.
Session 2: Computer-Assisted Interpreting (CAI) with a focus on ASR (Friday, 24th Jan @ 4pm GMT)
This session will discuss how Automatic Speech Recognition (ASR) technology and other CAI tools can support interpreters during real-time interpretation. It will examine research evidence to demonstrate the opportunities CAI tools can offer, especially those leveraging ASR, discuss what is necessary to exploit these opportunities effectively (including training and customisation), explore how CAI tools transform professional practice and workflows, and highlight pitfalls and risks for interpreters and organisations.
Session 3: LLMs to support interpreters (Wednesday, 29th Jan @ 5pm GMT)
This session will explore how LLMs can support interpreters when they prepare for an assignment. The session will discuss prompting of LLMs and give participants the opportunity to build glossaries with the help of LLMs. For this task we will use Google’s NotebookLM. In this session, we will also discuss how to improve the accuracy of an ASR engine with the help of the extracted terms, as a first step in the customisation process.
Session 4: Customising ASR with the help of a domain specific corpus. Discussion of risks. (Friday, 31st Jan @ 4pm GMT)
This session will discuss how to customise an ASR engine with the help of domain specific corpora. We will first discuss how we can collect relevant documents from the web to create a domain specific corpus. We will upload this corpus to a platform like Amazon transcribe to create a customised ASR engine. The second part of the session will wrap up the course with a discussion about the risks of using ASR and AI in general, in the context of interpreting.