Agentic Retrieval-Augmented Generation for Terminology Translation in Neural Machine Translation

This PhD project investigates terminology translation in neural machine translation, with a focus on improving the handling of domain-specific terms in translation workflows. The research addresses the problem that even advanced neural MT models and large language models still struggle to translate specialised terminology accurately and consistently in context. The project will explore methods such as: in-context learning with terminology examples glossary-constrained decoding agentic retrieval-augmented generation with term bases low-resource translation In the first year, the PhD candidate will focus on learning core NLP methods and text analytics techniques. In later stages, the candidate will define research objectives, develop innovations, and contribute to the state of the art. There may also be opportunities for collaboration with several research centres and partner institutions, including possible research activity abroad or in industry.

Honours degree, minimum 2:1, in Computer Science, Mathematics, Engineering, or a related technical discipline
Other relevant qualifications may also be considered

Desirable:

Master’s degree, preferably 2:1, in a relevant area
experience in NLP, computational social science, machine learning, or data science
experience with quantitative and qualitative research methods

Essential Knowledge and Skills:

strong Python programming skills
strong understanding of research philosophy, theory, and methodology
strong interest in NLP and machine translation
knowledge of qualitative and quantitative research methods
strong writing, presentation, and communication skills
ability to work independently and in a multidisciplinary team

Desirable Technical Skills:

NumPy, SciPy
Scikit-learn, Keras, TensorFlow
data analysis, visualisation, and interpretation

Application route is not shown in the text provided, so verify the official submission method before posting. Applicants whose first language is not English must meet SETU English language requirements and provide the required evidence with their application.