d-real is funded by Science Foundation Ireland and by the contributions of industry partners.
Applications for d-real positions starting in September/October 2022 are being accepted. We are operating a rolling call. Supervisors are currently assessing applications from the first and second calls. The third call is now open and will close at 16.00 (Irish time) on Monday 20th June 2022. If you have any questions about the programme please email stephen.carroll@d-real.ie. You can make an application here (through the Good Grants system). You will need to register (free) with Good Grants and confirm your email address to make an application. You will be asked to list your top three topic preferences (listed below). You will also be asked some personal details, your educational history, work experience, technical skill and for a statement on why you would like to join the programme. The system permits the uploading of 5 PDF documents, so you can upload supporting material (CV, personal statement, paper publications, etc. ).
Dublin City University
Code: 2022DCU01
Title: Using Disentangled Language Learning for Stylistically and Semantically Controllable Language Generation
Supervision Team: Anya Belz, DCU (Primary Supervisor) / Yvette Graham, TCD (External Secondary Supervisor)
Description:
Natural Language Generation (NLG) has made great strides recently owing
to the power of transformer language models (LMs). The best generators
produce text on a par with high-quality human writing, but have a very
high carbon footprint and are marred by hallucinated content, and
logical and factual errors. The general question addressed by this
thesis proposal is: to what extent can LM learning be
disentangled so that different aspects of language can be learnt and
retained separately, potentially
enabling (i) more direct control over meaning and style, (ii) different
systems to share generic
knowledge, and (iii) a system to generate the same meaning in any number
of different idiolects,
among other benefits, including a reduction in energy requirements. The
work will start with
experiments (i) generalising basic disentangling language-GANs, and (ii)
adapting state-of-the-art
text-to-image techniques capable of controlling style and semantics of
generated images via
autoregressive transformers.
Code: 2022DCU02
Title: Development of machine learning tools to predict pathological sequelae of traumatic
brain injury
Supervision Team: David MacManus, DCU
(Primary Supervisor) / Caitríona Lally, TCD (External Secondary
Supervisor) / David Loane, TCD (Additional Supervisory Team Member)
Description:
Traumatic brain injuries (TBI) are one of the leading causes of death
and disability worldwide.
Currently, there is a huge gap in diagnostic and prognostic technologies
for TBI likely due to its multiple
biological and biomechanical aspects which can cause different
neurological pathologies, impairments,
and deficits in people. This multifaceted nature of TBI makes it
difficult to predict the pathological
outcomes using existing technologies. Indeed, it remains difficult to
not only diagnose certain TBIs e.g.,
concussion, but it is also difficult to determine an accurate prognosis.
Therefore, the aim of this research
is to develop state-of-the-art computational tools to predict TBI
pathologies and provide accurate
diagnoses and prognoses. This will be achieved by using TBI pathology
data (e.g., MRI) with state-of-the-art computer models of the brain to
develop novel computational tools utilising machine learning to
determine accurate diagnoses and prognoses of TBI from head impacts
e.g., sports-related head impacts,
falls.
Code: 2022DCU03
Title: Autonomous Mapping and Exploration Using Unmanned Aerial Vehicle in GPS-Denied Environments
Supervision Team: Hossein Javidnia, DCU (Primary Supervisor) / Peter Corcoran, NUIG (External Secondary Supervisor)
Description:
As a requirement, majority of the Autonomous Aerial Vehicles (AAV) must
be provided with a prior well-constructed environment model. Such
prerequisite restricts the performance and applicability of the vehicle
and of course providing such a model for complex scenes is not always
feasible. This issue calls for the AAVs to perceive an unseen
environment in 3D and perform the desired navigation and manipulation
tasks. This issue becomes more significant in the absence of GPS
signals. Navigating through a GPS-denied environment requires the
vehicle to perceive the scene using a LIDAR and 3D sensing modules as
well as a complex path planning strategy which is recently being
investigated using ML and reinforcement learning methodologies.
Code: 2022DCU04
Title: Immersive Mathematics – Design for All
Supervision Team:
Tracey Mehigan, DCU (Primary Supervisor) / Emma Murphy, TU Dublin
(External Secondary Supervisor) / Dr Monica Ward, DCU (Additional
Supervisory Team Member)
Description: Mathematical
equations are an essential component of education for students across
all levels however they are not always presented to students in a manner
that suits their individual needs. For example, students with dyslexia
may struggle with symbols but benefit from spatial representations
whereas for blind students, spatial layout becomes a barrier. To address
these and other issues, this project explores the potential of
multimodal feedback in immersive environments for personalised
presentation of mathematics via an Artificial Intelligent Educational
(AIEd) framework for automatic adaptation to learner, capability and
learning situation.
Code: 2022DCU05
Title: How Full is Your Tank…Development of a ‘Readiness-to-Perform’ Tool for Monitoring & Predicting Sporting Performance
Supervision Team:
Anna Donnla O’Hagan, DCU (Primary Supervisor) / Jane Walsh, NUIG
(External Secondary Supervisor) / Sarah Meegan, DCU (Additional
Supervisory Team Member)
Description: The sports
industry is a multi-billion-dollar industry in which coaches and sport
and medicine teams strive to push and progress an athlete’s performance
year on year. Coaches and scientists structure training programmes with
distinct periods of progressive overload coupled with recovery in an
attempt to maximise or sustain performance during specific periods of
competition. The objective of this research is to test the most relevant
existing, portable, technologies to monitor athletes’ physical,
cognitive, and psychological readiness for sporting performance and to
develop a protocol combining bio-sensory and performance-based data
acquisition (e.g., mobile version of Psychomotor Vigilance Task, CANTAB,
wearable HRV/blood pressure devices, mobile EEG ECG etc.) related to
performance readiness to aid athletes and coaching personnel on an
athletes individualised readiness for sporting performance. The project
will ultimately deliver a proposal for a user-friendly tool for
monitoring performance readiness to assist with continuous improvement
in sporting performance and training practices.
Code: 2022DCU06
Title: Usable Interaction for Generic Menu Navigation in Virtual Reality
Supervision Team: Hyowon Lee, DCU (Primary Supervisor) / Brendan Rooney, UCD (External Secondary Supervisor)
Description:
There are many demonstration systems of VR technology in various fields
today, but there is a considerable lack of standardized, agreed-upon or
proven generic menu navigation interaction within a VR environment
resulting in poor usability and user-experience (UX) in most of these
demonstration systems. Without offering more usable menu interaction
through better understanding of the special characteristics of VR
interaction and the affordances it entails, VR devices and tech demos
have little chance of becoming a truly ubiquitous interaction platform
as its desktop and mobile counterparts. This project will develop
usability principles/design guidelines for VR interactivity especially
focusing on a generic menu navigation within the VR environment, with a
set of usable menu navigation strategies and accompanying suite of UIs
for them, both co-evolving and co-refined through a series of
design-centric iterations of design-prototyping-testing cycles.
National University of Ireland, Galway
Code: 2022NUIG01
Title: On-Device Neural Speech Understanding for consumer devices
Supervision Team:
Michael Schukat, NUIG (Primary Supervisor) / Naomi Harte, TCD (External
Secondary Supervisor) Peter Corcoran, NUIG, Gabriel Costache, Xperi,
Martin Walsh, Xperi (Additional Supervisory Team Members)
Description:
Today’s voice-based interfaces rely on a cloud-based infrastructure for
data processing and interpretation that causes issues with regard to
access to personal voice-data by large corporations. Therefore, there is
a growing trend in industry to move data-processing and analysis closer
to the source of data – the microphone that senses speech data. This
can be achieved using newly-developed neural accelerators (i.e. NVIDIA’s
Jetson, Google’s TPU, Xilinx Vitus-AI and Perceive’s Ergo) that
implement emerging neural-processing techniques.
This research aims to investigate emerging trends in speech analysis and
understanding with a focus on neural implementations and optimizations
for the above accelerator platforms. It will examine memory and
data-bandwidth aspects of recurrence in neural speech analysis, explore
neural speech enhancement techniques to pre-process voice signals that
are picked up from low-cost microphones, explore speech representations
for neural accelerator platforms, and deliver a proof-of-concept
smart-speaker, demonstrating feasibility of a practical stand-alone
neural speech interface. Previous project experience in digital speech
processing and neural network algorithm development is desirable.
Code: 2022NUIG02
Title: Executable Design: AI Tools for Music
Supervision Team: James McDermott, NUIG (Primary Supervisor) / Seán O’Leary, TU Dublin (External Secondary Supervisor)
Description:
This project is about new AI-enabled creative tools for musicians and
new creative ways for music consumers to interact with music. Modern AI
methods can generate plausible-sounding music, at least in some cases.
However, AI has not achieved anything like “understanding” of the
internal structures, relationships and patterns which are deliberately
designed by human composers. This understanding can best be represented
by short programs, since (by the Church-Turing thesis) no other
representation can be more powerful. Short programs are those which best
capture all possible regularity and structure.
We will use modern metaheuristic search methods such as genetic
programming, and neural program synthesis, to automatically create short
programs which, when executed, output given pieces of pre-existing
music. Simple manipulations of these programs will give natural
variations and extensions of the music, enabling exciting new tools for
creativity.
Code: 2022NUIG03
Title:
IntentPredictVR: A deep learning-based pedestrian intent prediction
from gaze, head, and body pose aided with synthetic data from VR
Supervision Team:
Ihsan Ullah, NUIG (Primary Supervisor) / Susan McKeever, TU Dublin
(External Secondary Supervisor), Peter Corcoran, Michael Schukat, and
Michael Madden (all NUIG)(Additional Supervisory Team Members)
Description:
Recent technology that works on deep neural networks has shown good
results for autonomous vehicles. Synthetic data generation from game
engines e.g. unreal and using it for training deep learning models where
data is scarce, has shown promising results in various applications.
Pedestrian protection is a critical challenge for these vehicles due to
pedestrian’s abrupt movement. For pedestrian safety, we will use a
similar technique in our system that will utilize pedestrian gaze, head
and body pose information from real and synthetic images to predict the
pedestrian intended direction. This plays a vital role in warning the
autonomous vehicle before approaching the pedestrian and avoiding
accidents, thus saving human lives.
Trinity College Dublin
Code: 2022TCD01
Title: Procedural Generation of Narrative Puzzles
Supervision Team: Mads Haahr, TCD (Primary Supervisor) / Marguerite Barry, UCD (External Secondary Supervisor)
Description: Narrative puzzles are puzzles that form part of the progression of a narrative, whose solutions involve
exploration and logical as well as creative thinking. They are key components of adventure and storydriven
games, and often feature in large open-world games. However, filling large open worlds with
engaging content is challenging, especially for games with procedurally generated worlds, such as
Minecraft (2011) and No Man’s Sky (2016). Systems exist for generating narrative puzzles
procedurally, but they lack context about many narrative elements, such as character motivation, plot
progression, dramatic arc, as well as player modelling. This project will improve procedurally generation
of narratives for small-scale narrative games as well as large-scale open world games by integrating new
types of narrative elements as well as player modelling into the Story Puzzle Heuristics for Interactive
Narrative eXperiences (SPHINX) framework, potentially resulting in dynamically generated narratives
of increased sophistication and significantly improved player experience.
Code: 2022TCD02 (no longer recruiting)
Title: Deep Learning for Magnetic Resonance Quantitative Susceptibility Mapping of Carotid Artery Plaques
Supervision Team:
Caitríona Lally, TCD (Primary Supervisor) / Catherine Mooney, UCD
(External Secondary Supervisor) / Alan Stone, St. Vincent’s Hospital
Dublin (Additional Supervisory Team Member)
Description:
Carotid artery disease is the leading cause of ischaemic stroke. The
current standard-of-care involves removing plaques that narrow a carotid
artery by more than 50%. The degree of vessel occlusion, however, is a
poor indication of plaque rupture risk, which is ultimately what leads
to stroke.
Plaque mechanical integrity is the critical factor which determines the
risk of plaque rupture, where the mechanical strength of this tissue is
governed by its composition. Using machine learning approaches and in
vivo imaging, and in particular Quantitative Susceptibility Mapping
metrics obtained from MRI, we propose to non-invasively determine plaque
composition and hence vulnerability of carotid plaques to rupture.
This highly collaborative project has the potential to change diagnosis
and treatment of vulnerable carotid plaques using non-ionizing MR
imaging which would be truly transformative for carotid artery disease
management.
Code: 2022TCD03
Title: Motion in the Metaverse: Perception of identity and personality from embodied humans
Supervision Team:
Rachel McDonnell, TCD (Primary Supervisor) / Brendan Rooney, UCD
(External Secondary Supervisor) / Aphra Kerr, TU Dublin (Additional
Supervisory Team Member)
Description: The Metaverse
is expected to be a digital reality combining social media and gaming,
with augmented and virtual reality, allowing users to interact virtually
for both work and entertainment. The Metaverse cannot exist without
avatars, which are virtual manifestations of the humans interacting in
the space. In this project, we will investigate the perception of motion
of avatars in social interactions in immersive Virtual Reality. In
particular, we are interested in how the quality of motion mapped from
the human affects perception of personality and personal identity.
Additionally, the ethical, legal and social implications around how
human motion data is captured and stored in the Metaverse will be
investigated as part of this project.
Code: 2022TCD04
Title: Personalisation of relapse risk in autoimmune disease
Supervision Team:
Mark Little, TCD (Primary Supervisor) / John Kelleher, TU Dublin
(External Secondary Supervisor) / Declan O’Sullivan, TCD and Alain
Pitiot (Ilixa software)(Additional Supervisory Team Members)
Description:
The PARADISE study targets development of a clinical decision support
tool that personalises immunosuppressive drug (ISD) therapy in
autoimmune disease. Using ANCA vasculitis as the exemplar condition and
leveraging off the Rare Kidney Disease registry and biobank, we will
focus on deep phenotyping of the patient in remission. At this time
point, we hypothesise that residual sub-clinical immune system
activation renders the patient at high risk of subsequent relapse of the
disease. Conversely, reversion of the immune system to a healthy
resting state may indicate a very low flare risk. By using a novel
semantic web technology, we will integrate clinical, patient app-derived
and multi-modal biomarker data streams to generate explainable machine
learning models that predict the risk of flare. These will inform the
physician about increasing ISDs or, indeed, discontinuing them
altogether. We envisage that this assessment will reduce both relapse
and ISD-associated infection, reduce healthcare costs, increase quality
of life and build human capital in a research area of importance to
Ireland.
Code: 2022TCD05
Title: Authenticity in Dialogue
Supervision Team: Carl Vogel, TCD (Primary Supervisor) / Eugenia Siapera, UCD, (External Secondary Supervisor)
Description:
Authenticity in communication is of utmost importance to those who
attempt to feign authenticity and is also relevant to those who would
prefer to banish inauthenticity, whether the sphere is public relations,
politics, health care, dating or courts of law. Dialogue interactions
in multiple modalities will be analyzed with the aim of identifying
features that discriminate authentic and pretended engagement. The work
will involve assembling a multi-modal corpus of evidently un-scripted
dialogue interactions, annotation with respect to authenticity
categories of interest, and analysis through combinations of close
inspection, semi-automated processing and data mining to identify
features that separate authentic and inauthentic dialogue
communications..
Code: 2022TCD06
Title: Automotive Audio System Integration and Tuning in Virtual Reality
Supervision Team:
Enda Bates, TCD (Primary Supervisor) / Andrew Hines, UCD (External
Secondary Supervisor) / Martin Walsh, Xperi/DTS (Additional Supervisory
Team Member)
Description: This project involves the
development of a VR system for automotive audio system integration and
tuning, in collaboration with Xperi/DTS. Modern automotive audio systems
are often acoustically tuned in-situ using a real physical car,
however, at this late stage of the design process, it may be impossible
to implement any significant modifications. Ideally, much of this
acoustic design could be done earlier in the design phase of the
automobile and the work done remotely or virtually, thereby potentially
greatly reducing the acoustics design cycle of a new car model, and
yielding cost savings to the manufacturer. This project will therefore
examine the use of VR for the accurate simulation of both the visual and
acoustic characteristics of automobiles and associated audio systems,
and compare the effectiveness of designing the acoustic system using the
traditional on-site approach with newly-developed remote simulation
methodologies based on this VR model.
Code: 2022TCD07
Title: When digital feels human: Investigating dialogue naturalness with multivariate neural data
Supervision Team: Giovanni M. Di Liberto, TCD (Primary Supervisor) / Benjamin Cowan, UCD (External Secondary Supervisor)
Description:
The interaction with digital systems has become a pervasive daily
experience (e.g., video-calls, dialogue systems). One major barrier
remains that users need to adapt the way they communicate to each
particular digital system, for example constituting a challenge for
inclusivity. This project will identify objective indices that quantify
the naturalness of a conversation by using bio-signals recorded with
electroencephalography and pupillometry. These metrics will inform us on
how exactly different digital communication strategies (e.g.,
video-call software) impact cognition (e.g., cognitive load,
phonological processing, temporal expectations). In doing so, this
project will inform us on the key elements for producing adaptive
dialogue systems.
Code: 2022TCD08
Title: An Intervention using a Personalized VR Dance/Game Tutor with Progress Monitoring for Survivors of Childhood Cancer
Supervision Team:
Juliette Hussey, TCD (Primary Supervisor) / Shirley Coyle, DCU
(External Secondary Supervisor) / Carol O’Sullivan, TCD (Additional
Supervisory Team Member)
Description: Recent
improvements in survival rates from childhood cancer have resulted in a
shift in focus, with more emphasis being placed on exploring and
reducing late-effects, addressing cognitive and physical rehabilitation
and enhancing the independence of long-term childhood CNS tumour
survivors. Known late-effects amongst survivors of childhood CNS tumours
include endocrinopathies, obesity, cerebrovascular complications,
cardiovascular conditions, neurologic conditions, neurocognitive
dysfunction and pulmonary dysfunction, all of which can negatively
impact on overall quality of life. In this project, we will design a
personalized intervention using a VR dance/game tutor for survivors of
childhood cancer. The intervention will be motivating and enjoyable for
the target population and will involve the use of VR headsets and
low/medium-cost tracking devices. The system will be personalized for
each participant and will be designed to include motivational and
monitoring features.
Code: 2022TCD09
Title: AI powered dynamic network control for real-time video streaming
Supervision Team: Marco Ruffini, TCD (Primary Supervisor) / Suzanne Little, DCU (External Secondary Supervisor)
Description:
This project aims to develop intelligent control mechanisms to make
network configuration decisions based on real-time information from live
video streaming and AR/VR services. The focus is on the cooperative
performance optimization for live video streaming and AR that considers
the interplay between service prioritization and resource availability
prediction when making control decisions. The goal is to make
predictions for video processing (e.g., encoding level, chunk size)
based on the anticipated capacity and latency in the network, which
depends on multiple environmental factors, including user behavior
(i.e., due to the highly interactive nature of the AR/VR content). On
the other hand, we will also make predictions about the network
performance and rely on various control loops, defined by the O-RAN
architecture, to dynamically reconfigure the network to match different
traffic requirements.
Code: 2022TCD10
Title: Designing apps to support psychotropic medication management and tapering
Supervision Team:
Gavin Doherty, TCD (Primary Supervisor) / Jane Walsh, NUIG (External
Secondary Supervisor) / Cathal Cadogan, TCD (Additional Supervisory Team
Member)
Description: Psychotropic medications, such
as antidepressants, are widely used in the treatment of mental health
disorders, and can be a valuable approach. However, evidence supporting
long-term use is often lacking and can be seen as inconsistent with a
recovery-oriented approach which emphasises service user engagement,
respect for autonomy and personhood. Previous research on technology
supports for people taking medication have not focussed on the use of
psychotropic medications, and the majority of the existing literature on
the topic has focussed on medication adherence. This PhD studentship
will examine the design of technology to support people taking
psychotropic medication, taking a more holistic view, looking at safe
and managed medication tapering/discontinuation as well as adherence and
management. It will involve work with multiple stakeholders, including
clinicians, user groups and people who use or have used these
medications. The PhD will be centered in Human-Computer Interaction,
with interdisciplinary input from a supervision team which involves
academics in Psychology and Pharmacy.
Technological University Dublin
Code: 2022TUD01
Title: Co-design of an interactive wellness park: A multimodal physical web installation for outdoor rehabilitation and wellness
Supervision Team:
Damon Berry, TU Dublin (Primary Supervisor) / Mads Haahr, TCD (External
Secondary Supervisor) / Emma Murphy, TU Dublin (Additional Supervisory
Team Member)
Description: Physical rehabilitation is
a critical and widely-applicable healthcare intervention. Increased
engagement in rehabilitation improves health outcomes and reduces
healthcare costs. Prescribed outdoor physical exercise can promote
social interaction and improve quality of life.
For this work, a co-designed physical web installation will be created
to make managed rehabilitation exercises more engaging and sustainable.
QR codes, NFC, and BLE will enable low-barrier connections to web
resources to support managed exercise. The user interface will be
co-created by service users and clinicians informed by behaviour change
theory to create a personalised and accessible outdoor digital
rehabilitation intervention.
The proposed system will comprise:
– Physical web installation residing on wooden posts in a healthcare campus.
– Web infrastructure enabling exercise regimes to be personalised.
Through co-design with stakeholders, physical web UX will be
investigated to assess different approaches in order to produce a
working installation and to develop an accessible and reproducible
design.
Code: 2022TUD02
Title: Controllable Consistent Timbre Synthesis
Supervision Team: Seán O’Leary, TU Dublin (Primary Supervisor) / Naomi Harte, TCD (External Secondary Supervisor)
Description:
The goal of the research will be to provide control over the design of
consistent musical instruments. Until recently sound synthesis has been
dominated by two approaches – physical modelling and signal modelling.
Physical models specify a source. Once the source is specified the
family of sounds coming from the source can be synthesised. Signal
models, on the other hand, specify waveforms and so are very general.
The major downside of signal models is that many parameters are required
to specify a single sound. The goal of this project is to use machine
learning algorithms to synthesise the parameters for a family of sounds
related to a single source. This project will marry machine learning and
signal processing techniques, including research into the use of
generative algorithms, signal models and sound representations.
Code: 2022TUD03
Title: Empowering older adults to engage with their own health data using multimodal feedback
Supervision Team: Emma Murphy, TU Dublin (Primary Supervisor) / Enda Bates, TCD (External Secondary Supervisor)
Description:
Health data from physiological sensors is often conveyed to users
through a graphical interface but this is not always accessible to
people with disabilities or older people due to low vision, cognitive
impairments or literacy issues. Real-time user feedback may not be
conveyed easily from sensor devices through visual cues alone, but
auditory and tactile feedback can provide immediate and accessible cues
from wearable devices. To avoid higher cognitive and visual overload,
auditory and haptic cues can be designed to complement, replace or
reinforce visual cues. This research will involve an exploration of the
potential of multimodal cues to enhance the accessibility of health
information from personal sensing devices used to monitor health
parameters such as blood pressure, sleep, activity, heart rate etc. By
creating innovative and inclusive user feedback it is more likely that
users will want to engage and interact with new devices and with their
own data.
Code: 2022TUD04
Title: Verbal Language to Irish Sign Language Machine Translation: A Linguistically Informed Approach
Supervision Team: Irene Murtagh, TU Dublin (Primary Supervisor) / Andy Way, DCU (External Secondary Supervisor)
Description:
Machine translation (MT) of verbal language (speech/text) has garnered
widespread attention over the past 60 years. On the other hand,
computational processing of signed language has unfortunately not
received nearly as much attention, resulting in its exclusion from
modern language technologies. This exclusion, leaves deaf and
hard-of-hearing individuals at a disadvantage, aggravating the
human-to-human communication barrier, while suppressing an already under
resourced set of languages further for the estimated 72 million deaf
people in the world. This aim of this project is to develop a
linguistically motivated sign language machine translation (SLMT) avatar
that will translate between English (text/speech) and Irish Sign
Language (ISL). The project will focus, in particular, on current
linguistic and technical challenges in relation the computational
modelling and processing of sign language. This will involve research in
sign language linguistics, computational linguistics, natural language
processing and virtual character animation.
Code: 2022TUD05
Title: Composing New Solutions to the Cocktail Party Problem: An Evolutionary Optimization Approach to Speech Reconstruction
Supervision Team: Ruairí de Fréin, TU Dublin (Primary Supervisor) / James McDermott, NUIG (External Secondary Supervisor)
Description:
The human brain can separate-out one speaker from a mixture of
speakers. This problem is commonly called the cocktail party problem.
Deploying signal enhancement techniques in assisted living environments
to enhance socialization opportunities for inhabitants is an open
problem. Techniques that can blindly separate an arbitrary number of
speech/music sources using a small number of anechoic mixtures, in real
time, could be used to profound effect in removing acoustic clutter in
these scenarios, however, the resulting separated speech does not sound
natural. This project will develop algorithms for composing natural
sounding speech for separated speakers using speaker-tuned fitness
functions and evolutionary optimization. The resulting algorithms will
be used to develop systems for natural speech production to encourage
social interaction amongst older adults.
Code: 2022TUD06
Title: Monitoring and Short-term Forecasting of Atmospheric Air Pollutants Using Deep Neural Networks
Supervision Team: Bianca Schoen-Phelan, TU Dublin (Primary Supervisor) / Soumyabrata Dev, UCD (External Secondary Supervisor)
Description:
Air pollution is a persistent problem in most of the world’s cities. It
has a significant negative influence on citizen health and quality of
life. Therefore, it is important to continuously monitor the pollution
concentrations and provide short-term forecasts. Historically, this
forecasting has been done quite poorly; traditional statistical
forecasting methods are unreliable for short-term predictions. Most
models are statistical and are limited in range of forecast time and
effectiveness. The goal of this PhD project is to create an intelligence
system that uses a combination of computer vision and deep learning
technologies to identify, monitor, and forecast air pollution in real
time, as well as provide residents with an early warning system. This
PhD project will also assess the key meteorological variables that
affect atmospheric air pollutant concentrations and examine the
forecasting model’s effectiveness for the island of Ireland,
particularly for the Dublin metropolitan area.
Code: 2022TUD07
Title: Computerized support for the Patient Generated Health Data lifecycle
Supervision Team:
Dympna O’Sullivan, TU Dublin (Primary Supervisor) / Lucy Hederman, TCD
(External Secondary Supervisor) / Damon Berry, TU Dublin (Additional
Supervisory Team Member)
Description: Patient
generated health data (PGHD) is increasing in importance with more
patients self-managing their conditions. Integrated PGHD including
enhanced computational support for acquisition, analytics and actions
can provide a holistic view and facilitate shared decision making. There
are numerous challenges – PGHD is large in volume and variety,
originating from medical devices, sensors and apps. Multiple standards
leads to semantic and syntactic incompatibilities. Sociotechnical
factors – adherence to self-management regimes, adoption of technology
and perceived trustworthiness of PGHD are important considerations. This
research will design, develop and evaluate a digital platform for the
PGHD lifecycle and focus on the following challenges;
– Understanding the requirements of patients and clinicians for self-management;
– Developing computerized support to capture and represent PGHD;
– Developing methods for analysis of PGHD that can support clinicians
from a remote management perspective while also supporting patients with
self-management;
– Evaluating the platform with patients and clinicians from HSE living labs.
University College Dublin
Code: 2022UCD01
Title: Data Aware Design: A framework for understanding human-data interaction in HCI and AI
Supervision Team:
Marguerite Barry, UCD (Primary Supervisor) / Aphra Kerr, TU Dublin
(External Secondary Supervisor) / David Coyle, UCD / Catherine Mooney,
UCD / Dave Lewis, TCD / Declan O’Sullivan, TCD (Additional Supervisory
Team Members)
Description: Many everyday digital
interactions are reliant on personal data being processed, often
passively and without user awareness, to allow for health tracking,
social media and digital content delivery, voice assistant services,
etc. The dependence of these applications on data and machine learning
has created an urgent need for data aware design in computer science
education, and in human computer interaction (HCI) and AI research. This
PhD project will investigate how application designers understand and
design for data use across services that use AI. Using qualitative
methods, including interviews with practitioners, it will explore
human-data interaction (HDI) and human computer interaction (HCI) theory
and methods for understanding data interactions. The aim is to develop
an interdisciplinary framework for data aware design in teaching and
research, to support understanding between and to promote human-centred
AI application design. This project would suit applicants interested in
qualitative research for understanding people and practices.
Code: 2022UCD02
Title: Supporting help-seeking and recommendations for mental health in young adults
Supervision Team:
David Coyle, UCD (Primary Supervisor) / Gavin Doherty, TCD (External
Secondary Supervisor) / Marguerite Barry, UCD and Claudette Pretorius,
UCD (Additional Supervisory Team Members)
Description:
Seeking help is a critical first step in addressing mental health
difficulties. Evidence suggests that positive help-seeking experiences
contribute to an increased likelihood of future help-seeking and to
improved mental health outcomes. Increasingly help-seeking now starts
online. However, help-seeking is a complex process. This project will
address known limitations of current online help-seeking technologies,
including a tendency towards information overload, medicalized
recommendations, and a lack of personalization. It will focus on the
help-seek needs of young adults, aged 18-25 and will be undertaken in
collaboration with national youth mental health organisations. The aim
is to develop guided help-seeking technologies including voice and
chat-based agent systems, social help-seeking technologies, and
conversational recommender systems. The research will be guided by past
research that has emphasised the importance of four key design
considerations: support for different levels of human connectedness,
accessible and trustworthy information, personalisation that respects
autonomy, and the need for immediacy. From a theoretical perspective it
will explore how traditional models of help-seeking can be integrated
with theories of information search and of engagement in Human Computer
Interaction.
Code: 2022UCD03
Title: Ethical AI: Developing principles and methods for ethical and trustworthy recommender systems
Supervision Team: Susan Leavy, UCD (Primary Supervisor) / Josephine Griffith, NUI Galway (External Secondary Supervisor)
Description:
AI driven recommender systems are becoming increasingly influential in
society. However, without appropriate regulatory frameworks, ethical
guidelines and auditing tools, they have the potential to promote
harmful content and lead to the perpetuation of social injustice.
Ethical issues presented by recommender systems concern issues such as
inappropriate content, risks to privacy and autonomy and lack of
algorithmic transparency (Milano et al., 2020). In response, regulations
are being developed nationally and within the EU and many public and
private organisations have published ethical guidelines for AI. This
research project aims to address the significant challenges in
translating policies and guidelines for ethical AI into implementable
design principles, methods and tools to enable audits and ethical
evaluation. This project will develop a theoretical framework for
ethical recommender systems, an audit methodology and tools to enable
systematic evaluation with particular consideration for the impact on
potentially vulnerable user groups.
Code: 2022UCD04
Title: VR for good: Exploring the active mechanisms underlying the use of Virtual Reality to prompt positive mood and well-being
Supervision Team: Brendan Rooney, UCD (Primary Supervisor) / Pamela Gallagher, DCU (External Secondary Supervisor)
Description:
Numerous studies report examples of virtual reality experiences being
used to prompt positive emotions and improve health and well-being. Yet
little is understood about the way in which such positive psychological
outcomes are designed into the virtual experiences – what are the active
mechanism by which such immersive experiences can bring about positive
emotions? This study explores the way in which the design of virtual
reality experiences interact with individual characteristics of the user
to impact perception, attention, emotion and mood, empathy, and
well-being. In order to do so, the study will identify and refine
effective measures to be used in the research (these may include using
self-report, physiological and cognitive performance measures). Then
building on the exploration of prospective mechanisms, the study will
field test some candidate experiences to prompt positive outcomes.
Code: 2022UCD05
Title: Imagining and Designing the Metaverse
Supervision Team:
Eugenia Siapera, UCD (Primary Supervisor) / Aphra Kerr, TU Dublin
(External Secondary Supervisor) / Cathy Ennis, TU Dublin (Additional
Supervisory Team Member)
Description: Interest in
the metaverse has increased dramatically since Mark Zuckerberg’s talk in
October 2021. By turning to the metaverse, Facebook/Meta indicated a
paradigm shift from a platform and social media based internet to an
immersive, integrated and experience-based environment. But what
precisely the metaverse will be is still undetermined, indicating that
the current period will shape its future form. It is therefore important
to study the sociotechnical imaginaries around the metaverse as they
will end up feeding into relevant policy and to the design of metaverse
applications. The project focuses on two key areas, games and health,
and seeks to identify the sociotechnical imaginaries of metaverse
applications in these areas as they are encountered among different
publics, including technology developers, gamers/users, and public
bodies. The project explores their views on challenges around user
engagement, privacy and other ethical issues, including transparency,
human dignity, individual and societal wellbeing, transparency,
accountability and non-discrimination. The outcome of the research is
expected to include a set of ethical and policy guidelines for the
metaverse.
Code: 2022UCD06
Title: Real-time Vision-based Product Placements in Multimedia Videos
Supervision Team: Soumyabrata Dev, UCD (Primary Supervisor) / Mimi Zhang, TCD (External Secondary Supervisor)
Description:
Product placement and embedding marketing are recently used extensively
for advertisement in today’s skip-ad generation. In this PhD project,
we use computer vision and deep learning techniques to accurately
perform product placement in multimedia videos. We intend to use
convolutional neural networks for accurately detecting existing adverts
in videos, tracking them across image frames, and replacing them with
new advertisements for targeted audiences. The designed neural networks
will be evaluated on available manually annotated data and synthetic
datasets. Such developed techniques will have wide-ranging impacts on a
variety of applications, including sports billboard marketing, retail
fashion advertising, and amongst others.
Code: 2022UCD07
Title: Perspective taking in multiparty human-machine dialogue
Supervision Team: Benjamin R Cowan, UCD (Primary Supervisor) / Vincent Wade, TCD (External Secondary Supervisor)
Description: Work on human-machine dialogue suggests that people take
their partner’s perspective into account when interacting with speech
based automated dialogue partners. Perspective taking more generally is
seen as critical to successful communication. Through smartspeakers,
speech agents have become devices that hold conversations with multiple
users in interaction. This move from dyadic to multiparty dialogue is
likely to be accelerated as agents take a more proactive approach to
engaging users in dialogue, becoming member of a team rather than the
sole target of the interaction. This PhD aims to identify how
perspective taking mechanisms manifest in mixed human-speech agent
teams, and how this influences user language choices when engaging with
the speech agent.
Link to original post at d-real website
