Don't miss our weekly PhD newsletter | Sign up now Don't miss our weekly PhD newsletter | Sign up now

  Exploring how Natural Language Processing (NLP) can help with the diagnosis of dementia

   Faculty of Engineering & Digital Technologies

  Dr Kulvinder Panesar,  Applications accepted all year round  Self-Funded PhD Students Only

About the Project

The Alzheimer’s Society acknowledges the COVID-19 pandemic led to a significant decrease in dementia diagnosis rates from 67.6% in February 2020 to 63.5% in June 2020. Here, primary care data showed a sharp drop of the number of referrals to the memory clinic. On average there are 2600 monthly referrals, yet these were lower in April (84) May, (435) and June (994) 2020 respectively. They further report “… referral numbers are increasing; a sustained and proactive effort must be made to support access to a timely diagnosis”. Dementia is a clinically complex, progressive neurological (brain) disorder. Alzheimer’s disease and vascular dementia are the most prevalent ("", 2020). Dementia symptoms include change of thinking speed, mental agility, language, understanding, judgement as well as memory loss, but each affected person will experience dementia differently. Dementia statistics from the NHS website state that there are over 850,000 people in the UK living with dementia; 7% of whom are over the age of 65, and the condition affects 17% of people over 80. It is estimated that by 2025 there will be over 1 million more people living with dementia.

The specific dementia symptoms of changes of language expression and comprehension, minor problems with cognition - MCI (mild cognitive impairment) and memory concerns are examples of triggers for a pro-active diagnosis, patient-specific prediction, evidence-based intervention, and  the need to support healthcare plans. 

Natural language (NL) is the most easily understood knowledge representation (KR) for people, but challenging for computers due to its inherent ambiguous, complex, and dynamic nature. The challenge of  natural language understanding (NLU) is demonstrated in the following example: (i) I am hungry; (ii) Is there any food in the fridge? (iii) My stomach feels empty; (i) to (iii) mean the same - I am hungry, as NL is highly variable, and our human mind, will understand the ideas behind each sentence (Panesar, 2020). Nuances of meaning make NLU difficult as the text’s meaning can be influenced by context and reader’s “world view” (Sharda, Delen, and Turban 2019). Natural language processing (NLP) is making sense of the data.

The 2020 pandemic has brought the future forward by five years. This is due to the AI adoption and the trend of moving from machine learning (ML) NLP problems of shallow models (e.g., SVM and logistic regression) trained on very high dimensional and sparse features to neural networks based on low dimensional, dense and distributed representations as deep new language models (DLM) demonstrating superior results and opportunities (Young et al., 2018).  

NLP can be beneficial  to diagnosing dementia in two ways.   Firstly, for the analysis of participant’s language and cognition tasks such as interviews or re-telling a story where the speech content can be analysed using innovative NLP using DLM, and pipelines. For example, ('MantrahLimited', 2020, Gorman, 2020), focused on healthcare support and early detection of dementia with smart devices.  Secondly, to address a range of  challenges. These are (i) NLU; (ii) the need for a  meaning representation of what is uttered; (iii )  to support the analysis results in (1)  – for a refined accuracy to any diagnosis.    One approach is  the psycholinguistic and cognitive adequacy (PCA) of functional models such as Role and Reference Grammar (RRG) (Van Valin Jr, 2005). PCA refers to psychological structures, principles and strategies which determine the way in which the linguistic expressions are acquired, generated, understood, processed, produced, interpreted, and stored in our in our mind (MAIRAL USÓN et al., 2019).

This project aims to develop hybrid (Ball, 2021b, Ball, 2021a, Saba, 2020, Mcshane and Nirenburg, 2021) – linguistic, ML and DLM experimental solutions for the diagnosis of dementia using evaluation benchmarks and liaising with the Centre for Dementia Studies for experimental work. A  small research group in NLP has published work on the motivations, design and evaluation of an NLP conversational agent – and is part of an international NLU and knowledge representation community. 

Applicants who have a background in computer science, NLP and an interest in language itself, with expertise in any of the following programming language (Python, Java, C, C++, or C#), and interested in a multidisciplinary development are welcome. The domain of this project can be adjusted as per the qualification and interests of students.

Computer Science (8) Languages, Literature & Culture (21) Linguistics & Classics (23)

Funding Notes

This is a self-funded PhD project; applicants will be expected to pay their own fees or have a suitable source of third-party funding. UK students may be able to apply for a Doctoral Loan from Student Finance for financial support.


'MANTRAHLIMITED'. 2020. Natural Language Processing based Knowledge Base and Chatbot for People with Dementia and Caregivers [Online]. UK Research and Innovation. Available: [Accessed 17th June 2021].
"ALZHEIMERS.ORG.UK". 2020. Alzheimer’s Society comment on how coronavirus is affecting dementia assessment and diagnosis [Online]. Available: [Accessed 9th March 2021].
"NHS.UK". 2020. About Dementia [Online]. NHS. Available: [Accessed 10th December 2020].
BALL, J. 2021a. Representing NLU with Meaning: Examples [Online]. "Pat.Inc". Available: [Accessed 3rd March 2021].
BALL, J. 2021b. Using Meaning as Universal Knowledge Representation [Online]. "Pat.Inc". Available: [Accessed 2 Mar 2021].
GORMAN, R. 2020. Early Detection of Dementia with Smart Devices - Digital biomarkers of cognitive decline could alert us to the early stages of dementia before irreversible damage occurs. The
MCSHANE, M. & NIRENBURG, S. 2021. Linguistics for the Age of AI, MIT Press.
PANESAR, K. 2020. Conversational artificial intelligence-demystifying statistical vs linguistic NLP solutions. Journal of Computer-Assisted Linguistic Research, 4, 47-79.
SABA, W. 2020. Why Ambiguity is Necessary, and why Natural Language is not Learnable [Online]. Available: [Accessed 12th January 2021].
VAN VALIN JR, R. D. 2005. A summary of Role and reference Grammar. Role and Reference Grammar Web Page, University of Buffalo.
YOUNG, T., HAZARIKA, D., PORIA, S. & CAMBRIA, E. 2018. Recent trends in deep learning based natural language processing. ieee Computational intelligenCe magazine, 13, 55-75.

Register your interest for this project