Engineering and Physical Sciences Collaborative Research Scholars

EPSRC collaborative studentship: Natural Language Processing for Improving Support for Young People’s Mental Health (2025)

What you get

As an EPSRC student you will join a vibrant doctoral community and will benefit from:

• 3.5 years of funding with a a tax free living allowance at the standard Research Council rate – currently £19,237 in 2024-5 – and international or UK PhD fees
• An option of a paid placement with the company
• Funding for training and research expenses, such as conference trips and experimental costs
• Supervision by world-leading researchers
• Our Entrepreneurship Summer School and Responsible Research and Innovation workshops tailored to the needs of engineering and physical sciences researchers.
• Our Researcher Development workshops and access to taught modules relevant to your project.

Type of award

Postgraduate Research

PhD project

This project will be in collaboration with  Tellmi,  an organisation specialising in providing digital mental health support for young people.  Their app provides a safe space for young people to talk about their experiences; pre-moderated, anonymous peer support; pre-emptive counsellor intervention and access to a directory of specialist resources.  Previous collaborations with Tellmi have focussed on using Natural Language Processing to support the moderation process as well as finding and recommending resources within the directory based on user posts.  Now Tellmi would like to investigate using NLP technology for directly answering user questions given certain constraints: user data (including posts) is highly sensitive making certain cloud services off-limits, building and running large language models locally is prohibitively expensive, users are vulnerable and may be at high risk if given incorrect information.  Given these constraints, how much can NLP and other AI technology be used to speed up response times in a scalable way?

Recent innovations in Natural Language Processing, including Large Language Models (LLMs), have already made it possible to create AI chatbots which can engage in conversation and answer questions.  However, due to their nature, large language models have a tendency to “hallucinate” and make things up.  Responses also tend to be very fluent and persuasive which make it hard for humans (and technology) to identify the incorrect information.  Some progress has been made with Retrieval Augmented Generation (RAG) which encourages the LLM to base its answer on trusted text, as well as prompt engineering and self-reflection / evaluation.  It is expected that this project might investigate some of these techniques and the extent to which they can mitigate against hallucinations.

This project is quite wide in possible scope and it is expected that a successful applicant will write their own project proposal which demonstrates their understanding of possible research questions and relevant literature in the field of NLP, as well as suggesting a possible project plan.

It is expected that the PhD work will focus on one or more of the research  questions from an academic perspective.  Methods will be developed and tested within the academic setting.  Other datasets as well as data provided by the project partner, Tellmi, will be used in training and evaluation.  Tellmi are offering a paid placement (which could be done on-block or part-time concurrently with the PhD) which would focus on software development and integration within their company systems.

Eligibility

Applicants should have/expect to have at least a 2:1 undergraduate honours degree in a relevant subject (e.g. Computer Science) and meet our English language requirements. International and UK applicants are both eligible to apply, but in line with EPSRC regulations, international awards will be limited.

Sussex recognises the value of diversity in research communities and is committed to addressing patterns of under-representation. We recognise that a range of factors, both structural and individual, influence and impede pathways into postgraduate research in different ways.  If you wish, you may write a short, optional statement offering any further contextual information affecting your pathway to postgraduate research not included elsewhere that you think is relevant to your application.

Number of scholarships available

One

Deadline

14 February 2025 16:00

How to apply

1. Contact the main supervisor to find out more about their research and what they are looking for in a PhD application.
2. Read our guidance on PhD applications and apply for a PhD place at Sussex using the online PhD application system.
3. Apply for the PhD course in INFORMATICS.  List Prof Julie Weeds as your preferred supervisor and enter “EPSRC iCASE 2025” in the “sources of funding” box.
4. Complete the EPSRC application form (in FILES section at the bottom of this page)  and upload it to your PhD application in the “Research Proposal “category. You may also upload an optional Contextual Information form (below)  if you wish to tell us more about any barriers you have faced. If the forms do not download with a Chrome browser, please use Edge or Firefox.
5. Upload as attachments along with your EPSRC application form, your CV. degree certificates and transcripts and English language qualifications if applicable before submitting the PhD application online.
6. Complete our online data collection form. This is not part of the assessment process but used for reporting to the EPSRC.

   https://qualtricsxm4dz52nmjl.qualtrics.com/jfe/form/SV_9FZsYtb6a855Ado

Relevant papers are:

1. Rummer-Downing, T. and Weeds, J. (2023). Leveraging Out-of-the-Box Retrieval Models to Improve Mental Health Support. In Proceedings of the 16th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2023) - HEALTHINF; ISBN 978-989-758-631-6; ISSN 2184-4305, SciTePress, pages 64-73. DOI: 10.5220/001163430000341
2. Wang, D. ; Weeds, J. and Comley, I. (2020). Improving Mental Health using Machine Learning to Assist Humans in the Moderation of Forum Posts. In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - HEALTHINF; ISBN 978-989-758-398-8; ISSN 2184-4305, SciTePress, pages 187-197. DOI: 10.5220/0008988401870197
3. Peng, Q. Weir, D.and Weeds, J. 2023. Testing Paraphrase Models on Recognising Sentence Pairs at Different Degrees of Semantic Overlap. In Proceedings of the 12th Joint Conference on Lexical and Computational Semantics (*SEM 2023), pages 259–269, Toronto, Canada. Association for Computational Linguistics.
4. Hunter, S.B., Mathews, F. and Weeds, J. 2023.Using hierarchical text classification to investigate the utility of machine learning in automating online analyses of wildlife exploitation.  Ecological Informatics, Volume 75, 2023, 102076, ISSN 1574-9541, https://doi.org/10.1016/j.ecoinf.2023.102076.

Sponsors

These scholarships are funded by the UK Engineering and Physical Sciences Research Council (EPSRC)

Contact us

For questions regarding your chosen research project, please contact the project supervisor: Professor Julie Weeds

If you have questions relating to the online application process, please contact enginf-pgr@sussex.ac.uk  for the School of Informatics and Engineering

You might also be interested in

  This project is part of the Sussex EPSRC Doctoral Landscape Award.  Further details are here:  

https://www.sussex.ac.uk/study/fees-funding/phd-funding/view/1818-EPSRC-Science-and-Engineering-studentships

Timetable

14^th February 2025 (23:59): Deadline for applications
January to March 2025: Interviews 
From late March 2025: Awards confirmed