Please note that this seminar confirmed for Purdie C on Wednesday October 4th between 14:00 and 15:00
Title: New file metaphors for a networked world
Continue reading
Abstract:
Natural language processing (NLP) is a multidisciplinary field closely related to linguistics, machine learning and artificial intelligence. It comprises a number of different subfields dealing with different kinds of analysis and/or generation of natural language texts. All these methods and approaches need some kind of evaluation, i.e. comparison between the obtained result with a given gold standard. For tasks dealing with text generation (such as speech recognition or machine translation), a comparison between two texts has to be carried out. This is usually done either by counting matched words or word sequences (which produces a similarity score) or by calculating edit distance, i.e. a number of operations needed to transform the generated word sequence into a desired word sequence (which produces a “dissimilarity” score called “error rate”). The talk will give an overview of advantages, disadvantages and challenges related to this type of metrics mainly concentrating on machine translation (MT) but also relating to some other NLP tasks.
Speaker bio:
Maja Popović graduated at the Faculty of Electrical Engineering, University of Belgrade and continued her studies at the RWTH Aachen, Germany, where she obtained her PhD with the thesis “Machine Translation: Statistical Approach with Additional Linguistic Knowledge”. After that, she continued her research at the DFKI Institute and thereafter at the Humboldt University of Berlin, mainly related to various approaches for evaluation of machine translation. She has developed two open source evaluation tools, (i) Hjerson, a tool for automatic translation error classification, and (ii) chrF, an automatic metric for machine translation evaluation based on character sequence matching.
Room 222 – Physics and Astronomy
12:05 Michael Mazilu: Introduction
12:15 Malte Gather and Nils Kronenberg: Developing cell forces mapping for clinical diagnosis
12:45 Vivienne Wild and Milena Pawlik: Analysing images of galaxies
13:15 Coffee Break
13:25 David Harris-Birtill : Automated Remote Pulse Oximetry
Talk Title: Monads and Lenses
Abstract:
Monads are an abstraction that can be used to mathematically model computational effects (among other things). Lenses are an abstraction for bidirectional computation, a generalization of the view-update problem. In this talk I will discuss ways to combine them and why it might be interesting to do so.
This talk is on joint work with Faris Abou-Saleh, Jeremy Gibbons, James McKinna and Perdita Stevens conducted as part of the recently-concluded project “A theory of least change for bidirectional transformations”.
Please note that this seminar will now take place in Jack Cole 1.33A on Wednesday 5th July between 15:00 and 16:00
Title: Between Exploration and Explanation: Visualizations for Insights, Curiosity, and Storytelling.
Abstract: This talk presents a set of interactive visualizations for exploration and recent work in how to communicate insights through data-driven stories. In particular, I will present work on visualizing networks including an open-source online platform. Then, I will discuss comics as an approach to communicate not only changes in temporal data but to weave narration, textual explanations, and data visualizations. The questions raised by the talk are about effective ways to engage a larger audience in understanding, learning, and use of visualizations for exploration and communication. As visualizations are becoming more and more commonplace and familiar to people, we can see more and more aspects of our daily lives being potentially enriched with information presented visually. Eventually, I want to raise the question of which role novel technology such as Augmented and Virtual Reality can play in exploring, communicating, and interacting with visualizations.
Biography: Benjamin is a Lecturer in Design Informatics and Visualization at the University of Edinburgh. His research designs and investigates interactive information visualizations to help people explore, present, and understand information hidden in data. He focuses on the visualization of dynamic networks (e.g., social networks, brain connectivity networks), as well as temporal data (e.g., changes in videos and Wikipedia articles, events on timelines), comics for storytelling with visualizations, as well as visualization and interaction in Augmented and Virtual Reality. Before joining the University of Edinburgh in 2017, Benjamin worked as a postdoc at Harvard University, Monash University, as well as the Microsoft-Research Inria Joint Centre. Benjamin was visiting researcher at the University of Washington and Microsoft Research in 2015. He obtained his PhD in 2014 from the Université Paris Sud where he worked at the Aviz Group at Inria.
Title: Finding What to Read: Visual Text Analytics Tools and Techniques to Guide Investigation
Abstract: Text is one of the most prominent forms of open data available, from social media to legal cases. Text visualizations are often critiqued for not being useful, for being unstructured and presenting data out of context (think: word clouds). I argue that we should not expect them to be a replacement for reading. In this talk I will briefly discuss the close/distant reading debate then focus on where I think text visualization can be useful: hypothesis generation and guiding investigation. Text visualization can help someone form questions about a large text collection, then drill down to investigate through targeted reading of the underlying source texts. Over the past 10 years my research focus has been primarily on creating techniques and systems for text analytics using visualization, across domains as diverse as legal studies, poetics, social media, and automotive safety. I will review several of my past projects with particular attention to the capabilities and limitations of the technologies and tools we used, how we use semantics to structure visualizations, and the importance of providing interactive links to the source materials. In addition, I will discuss the design challenges which, while common across visualization, are particularly important with text (legibility, label fitting, finding appropriate levels of ‘zoom’).
Biography: Dr. Christopher Collins is the Canada Research Chair in Linguistic Information Visualization and an Associate Professor of Computer Science at the University of Ontario Institute of Technology (UOIT). His research focus is interdisciplinary, combining information visualization and human-computer interaction with natural language processing to address the challenges of information management and the problems of information overload. His work has been published in many venues including IEEE Transactions on Visualization and Computer Graphics, and has been featured in popular media such as the Toronto Star and the New York Times Magazine. He received his PhD in Computer Science from the University of Toronto. Dr. Collins is a past member of the executive of the IEEE Visualization and Graphics Technical Committee and sits on the IEEE VIS Conference Organizing Committee.
“Addiction”
Seminar Room 1 School of Medicine
12:00: Alex Baldacchino- Introduction
12:15: Ognjen Arandjelović & Aniqa Aslam- Understanding Fatal and Non-Fatal Drug Overdose Risk Factors in Fife: Overdose Risk (OdRi) tool
12:45: Damien Williams & Fergus Neville- Transdermal alcohol monitoring
13:15: David Harris-Birtill & David Morrison- Narco Cat – waste water analysis in substance misuse – a novel epidemiological tool
13:15 – 14.00: All Questions & Opportunities
Event Location: School of Medicine, Seminar room 1
Abstract:
Session types are a formalism to model structured communication-based programming. A session type describes communication by specifying the type and direction of data exchanged between two parties. We show that session types are encodable in more primitive and foundational pi-calculus types. Besides providing an expressivity result, the encoding: (i) removes redundancies in the syntax of session types, and (ii) yields standard properties of session types as straightforward corollaries, exploiting the corresponding properties of standard typed pi-calculus. The robustness of the encoding is tested on a few extensions of session types, including subtyping, polymorphism, and higher-order communications. In this talk we present the encoding, some of its applications and recent developments.
Title: New Approaches to Mode-Switching
Abstract: The fundamental unit of all interaction is issuing commands, and the trickiest types of commands are those that control “modes” — different ways to map the same input to different application actions. For example, the current mode in a tablet drawing app could determine if the exact same sequence of touch movements draws a line, pans the canvas, makes a marquee selection, or issues a gestural command. Switching between modes like these are frequent, so finding optimum mode-switching methods is important. In this talk, I survey my group’s work to understand and improve mode-switching and command selection for different input types and device form factors. These include: Pin-and-Cross, a touch overloading technique combining static touches with nearby crossing selection; Conté, a pen-like input device that leverages small changes in contact geometry; Doppio, a reconfigurable two-faced smartwatch for tangible input; and Gunslinger, a mid-air interaction technique using bare hand postures and gestures performed in a relaxed arms-down position.
Bio: Daniel Vogel is an Assistant Professor and co-director of the Human-Computer Interaction Lab in the Cheriton School of Computer Science at the University of Waterloo. His research interests are fundamental characteristics of input and novel forms of interaction for current and future computing form factors like touch, tangibles, large displays, mid-air gestures, and whole-body input. In addition to earning PhD and MSc degrees from the University of Toronto, Dan holds a BFA from the Emily Carr University of Art + Design, and he leverages his combined art and science background in his research. For example, he was recently awarded a major grant to build a $1.8 million lab to explore the intersection of HCI and Fine Art in spatial augmented reality. Dan’s 2004 paper on interactive ambient displays is one of the ten most cited papers in the nineteen-year history of ACM UIST, and he has received honours including: multiple best paper awards at ACM CHI; the Bill Buxton Dissertation Award (2010); a Banting Postdoctoral Fellowship (2011 – 2013), and an Ontario Early Researcher Award (2017).
Session types codify communication patterns, giving developers guarantees that applications satisfy predefined protocols. Session types have come a long way from their theoretical roots: recent work has seen the implementation of static analysis tools; embeddings into a multitude of programming languages; and the integration of session types into languages as a first-class language construct.
Work at Edinburgh has concentrated on the latter. Lindley and Morris have extended the experimental functional programmming language Links with session-typed hannels in a multithreaded setting.
Distribution, however, brings challenges such as failure and the need for distributed channel delegation algorithms. In this talk, I will demonstrate and discuss the design and implementation of session types in Links. I will describe my recent work on adding support for distribution to Links, allowing the creation of session-typed, multi-user web applications.
Finally, I will describe recent, in-progress, work on a static type system and semantics allowing the controlled relaxation of the requirement of *linearity* (that
every endpoint must be used exactly once) to that of affinity (that every endpoint must be used at most once) in order to account for the question of users leaving a session midway through, and describe how the system still retains core metatheoretic pro
