We investigated how users evaluate passage-length answers for non-factoid questions. We conduct a study where answers were presented to users, sometimes shown with automatic word highlighting. Users were tasked with evaluating answer quality, correctness, completeness, and conciseness. Words in the answer were also annotated, both explicitly through user mark up and implicitly through user gaze data obtained from eye-tracking. Our results show that the correctness of an answer strongly depends on its completeness, conciseness is less important. Analysis of the annotated words showed correct and incorrect answers were assessed differently. Automatic highlighting helped users to evaluate answers quicker while maintaining accuracy, particularly when highlighting was similar to annotation. We fine-tuned a BERT model on a non-factoid QA task to examine if the model attends to words similar to those annotated. Similarity was found, consequently, we propose a method to exploit the BERT attention map to generate suggestions that simulate eye gaze during user evaluation.
Baranova Valeriia, PhD candidate at RMIT University (supervised by Mark Sanderson, Falk Scholer and Bruce Croft). The former head of research and development NLP department at Tinkoff Bank.
Systems Seminar: Optimized Contextual Data Offloading in Mobile Edge Computing
Mobile Edge Computing (MEC) is a new computing paradigm that moves computing resources closer to the user at the edge of the network. The aim is to have low-latency, high bandwidth, and to improve energy consumption when running computational tasks. The idea of deploying MEC servers near to the users along the 5G technology has led to open an interest in the field of Vehicular Network (VN). MEC servers can play significant roles in improving the performance of VN applications. In this environment, offloading computational tasks over collected contextual data by the mobile nodes (smart vehicles for example) meets the challenge of when & where to offload the collected data while on the move. In this work, we modelled the problem of offloading contextual data to the MEC servers as an optimal stopping problem. Our objectives are to offload to a MEC server with lower execution time and before the collected data get stale. We evaluated our model using real mobility trace with real servers’ utilization; the results showed that the proposed model outperforms other offloading methods.
Systems Seminar: Towards QoS-aware Provisioning of Chained Virtual Security Services in Edge Networks
Future networks are expected to deliver low-latency, user-specific services in a flexible and efficient manner. Operators have to ensure infrastructure resilience in the face of such challenges, while maintaining service guarantees for subscribed users. One approach to support emerging use cases is through the introduction and user of virtualised network functions (VNFs) at the edge of the network. While placement of VNFs at the network edge has been previously studied, it has not taken into account services comprised of multiple VNFs and considerations for network security. In this paper we propose a mathematical model for latency-optimal on-path allocation of VNF chains on physical servers within an edge network infrastructure, with special considerations for network security applications and operator's best practices. We acknowledge the challenges of employing optimal solutions in real networks and provide the Minimal Path Deviation Allocation algorithm for placement of security-focused network services in a distributed edge environment, minimising end-to-end latency for users. We then evaluate our placement results over a simulated nation-wide network using real-world latency characteristics. We show that our placement algorithm provides near-optimal placement, with minimal latency violations with respect an optimal solution, whilst offering robust tolerance to temporal latency variations.
Multiparty Session Types for Safe Runtime Adaptation in an Actor Language
Next week (27th April), I will give a PLUG talk on a recent ECOOP'21 paper (joint with Paul Harvey, Ornela Dardha, and Simon Gay), entitled Multiparty Session Types for Safe Runtime Adaptation in an Actor Language.
Human fallibility, unpredictable operational environments, and the heterogeneity (and corresponding resource constraints) of hardware devices are driving in the need for software to be able to adapt as seen in the Internet of Things or national telecommunication networks. Unfortunately, mainstream programming languages do not readily allow a software component to sense and respond to its operating environment, by discovering, replacing, and communicating with other software components that are not part of the original system design, while maintaining static correctness guarantees. In particular, if a new component is discovered at runtime, there is no guarantee that its communication behaviour is compatible with existing components.
We address this problem by using multiparty session types with explicit connection actions, a type formalism used to model distributed communication protocols. By associating session types with software components, the discovery process can check protocol compatibility and, when required, correctly replace components. Moreover, the use of session types throughout the software system design guarantees the correctness of all communication, whether or not it is adaptive.
We present the design and implementation of EnsembleS, the first actor-based language with adaptive features and a static session type system. We apply it to a case study based on an adaptive DNS server. Finally, we formalise the type system of EnsembleS and prove the safety of well-typed programs, making essential use of recent advances in non-classical multiparty session types.
Date and Time: Tuesday 27th April, 3PM
Due to the ongoing impact of the COVID-19 pandemic, PLUG talks will take place online via Zoom. The details will be sent via the PLUG mailing list and the SPLS Zulip instance nearer the time.
Abstract: This talk will first provide an overview of the claimed expectations from machine learning and AI in medicine. It criticises the misconceptions around personalised medicine in the context of multi-cause conditions such as cardiovascular diseases. It then provides a general overview of the machine learning research at Deep Medicine with examples of how this is guiding research and could lead to better decision making in clinical practice. Examples will primarily focus on the application of a range of machine learning models to complex Electronic Health Records for risk prediction, co-morbidity and patient clustering, risk factor identification, causal inference and model interpretability.
Kazem Rahimi is a Professor of Cardiovascular Medicine and Population Health, at the University of Oxford and a consultant cardiologist at the Oxford University Hospitals NHS Trust. His research interests include hypertension, heart failure, multimorbidity and cardiovascular risk management, using a variety of methodologies such as individual-patient meta-analysis, large-scale decentralised clinical trials, and digital health technologies.
Kazem leads the Deep Medicine programme at the Nuffield Department of Women’s and Reproductive Health with a major interest in the application of machine learning approaches to electronic health records. He also leads the Blood Pressure Lowering Treatment Trialists’ Collaboration (BPLTTC), which is an international collaboration of all the major trials of blood pressure lowering drugs. He is the Director of the Martin School Programme on Informal Cities and a Co-Investigator of the PEAK-Urban programme. ----
Dr. Mohammadhossein Mamouei is a Machine Learning Scientist at the University of Oxford. As a member of the interdisciplinary research group, Deep Medicine led by Prof Kazem Rahimi, Mo specialises in the applications of machine learning on large-scale electronic health records to extract new insights and develop models for prognostication and diagnostication. He is involved in the PEAK Urban and Informal Cities programmes where he focuses on the effects of environmental and sociodemographic factors on health outcomes.
Before joining the University of Oxford, he was a postdoctoral researcher at City, University of London focusing on the analysis of high-dimensional optical spectra and bio-signals, and a research assistant at University of Southampton focusing on urban mobility.
Mo's educational background is in Applied Mathematics-Systems and Modelling (PhD, City, University of London), and Electrical Engineering (MSc, City, University of London & BSc, Iran University of Science and Technology).
He also leads a team of data scientists at a charity, Apart of Me, that has embarked upon the ambitious project of delivering the first mobile game to help children cope with bereavement.
The educational landscape has seen a major shift over the past 12 months. As the world starts to heal, it is important to reflect on what worked and what did not, so we can provide new insights into ways of helping both students and educators move forward and ensure efficacy in teaching/learning. SICSA Education will host an online event on the 29th April, presenting a number of talks on lessons learnt from this agile shift from across the SICSA Education network. You can register to attend this event via the Eventbrite link.
It's just a few weeks until CHI 2021, and in keeping with tradition, the SICSA HCI theme is organising a Pre-CHI day to celebrate this year's contributions from SICSA-affiliated authors. The event will take place online, and will have two parts -- registered participants can attend either or both parts as they are able. 09:00 – 10:30 – CHI 5-minute video watch party We will curate a video containing as many of the CHI papers that feature SICSA affiliated authors as possible. The video will premiere live on YouTube for members of the community to watch together and discuss, and will remain available after the event. 10:30 – 12:30 – CHI Discussion, Methods in the Madness. We fully recognise that this year has been incredibly difficult for authors to carry out publishable work, and also collaboratively work with others in achieving this. Following the live video showcase, we will share best practice from successful authors, through a guided discussion with the SICSA community that will focus on the following: Recruiting participants Study adjustment for COVID The remote writing process The remote rebuttal process Any other area related to the HCI publication process that has changed over the last year Register for the event. Note that SICSA has also funded a limited number of Uber Eats vouchers available for those who want to have the full Pre-CHI day experience -- these will be distributed free of charge on a first-come first-served basis to SICSA members who choose that option on the registration form. We hope to see you all there on the 6th!
Organised by staff and PhD students, the First University of Glasgow Computational Biology Conferences aims to bring together researchers across the University of Glasgow who are interested or work in the field of computational biology. Topics of interest including but not limited to: computational approaches in omics, systems biology, population genetics, medical imaging, public health and clinical data. The conference will take place in the afternoons of 27th and 28th of May. We are pleased to confirm our keynote speakers: Prof Robert Insall, Dr Breda Cullen, Dr Richard Reeve, and Dr Mayetri Gupta.
We welcome abstract submissions from early career researchers and PhD students. Abstract deadline is April 30th, 2021.Abstract submission form available on our website: https://cbc.dcs.gla.ac.uk/
As a child, she dreamed of becoming a professional cake decorator. However, it soon became clear that her future was more likely to lie in mathematics.
She went on to obtain an MMath in Mathematics and Computer Science at the University of Oxford (Merton College) in 2009; She stayed in Oxford, working under the supervision of Alex Scott, to complete my DPhil in Mathematics (with a thesis entitled Graph Colourings and Games) in 2013. From 2012 to 2014, she was a Postdoctoral Research Assistant in the School of Mathematical Sciences at Queen Mary University of London, working with Mark Jerrum on the EPSRC-funded project Computational Counting.
She came to the University of Glasgow in 2014, where she held a lectureship in the School of Mathematics and Statistics for two years. In October 2016 she moved to the School of Computing Science, where she currently hold a five-year Royal Society of Edinburgh Personal Research Fellowship (funded by the Scottish Government).
International Workshop on Artificial Intelligence and Mental Health
The World Health Organisation states that mental health problems costs the global economy one trillion US dollars each year, whereas in UK, mental health services report that 19.7% of people over 16 years old show symptoms of anxiety and depression. The latest interactive technologies of social robots and virtual reality, powered by Artificial Intelligence (AI) promise new approaches for clinical treatment of psychiatric, developmental and cognitive problems. This workshop brings together experts in AI, computing and mental health, from both academia and industry, that work towards the development of new, computing based methodologies for detection, treatment and analysis of mental health issues. Besides providing an extensive overview of the state-of-the-art in the domain, the workshop aims at helping young researchers (including PhD students and postdocs) and raise awareness of multidisciplinary opportunities of computing and medical sciences in mental health.