School Seminar ‘Paraphrase Generation from Latent-Variable PCFGs for Semantic Parsing’ by Shashi Narayan

Abstract:

One of the limitations of semantic parsing approaches to open-domain question answering is the lexicosyntactic gap between natural language questions and knowledge base entries — there are many ways to ask a question, all with the same answer. In this paper we propose to bridge this gap by generating paraphrases to the input question with the goal that at least one of them will be correctly mapped to a correct knowledge-base query. We introduce a novel grammar model for paraphrase generation that does not require any sentence-aligned paraphrase corpus. Our key idea is to leverage the flexibility and scalability of latent-variable probabilistic context-free grammars to sample paraphrases. We do an extrinsic evaluation of our paraphrases by plugging them into a semantic parser for Freebase. Our evaluation experiments on WebQuestions benchmark dataset show that the performance of the semantic parser significantly improves over strong
baselines.

Bio:

Shashi Narayan is a research associate at School of Informatics at the University of Edinburgh. He is currently working with Shay Cohen onthe problems of spectral methods for parsing and generation. Before,he earned his doctoral degree in 2014 from Université de Lorraine,under the supervision of Claire Gardent. He received Erasmus MundusMasters scholarship (2009-2011) in Language and CommunicationTechnology (EM-LCT). He did his major in Computer Science and Engineeringfrom Indian Institute of Technology (IIT), Kharagpur India. He is interested in the application of syntax and semantics to solvevarious NLP problems, in particular, natural language generation,parsing, sentence simplification, paraphrase generation and questionanswering.

Event details

  • When: 26th January 2016 14:00 - 15:00
  • Where: Cole 1.33
  • Series: School Seminar Series
  • Format: Seminar, Talk

Seminar: “Data Exploration on Smart watches” by Dr Rachel Menzies

rachel menzies

Abstract:

For many of us, interacting with data on mobile devices such as phones and tablets is commonplace in our lives, e.g. phone call data, TV guide, maps, fitness and wearable data. With the introduction of smart watches, the screen size of mobile devices has dramatically decreased. This reduction in screen real estate provides challenges for the design of interfaces, including the presentation and exploration of data visualisations. Using bar charts as an example, this presentation will explore the shortcomings of current zooming techniques on very small screens and consider proposed guidelines for the development of simple data exploration applications. Key design features such as the need for overview and context will be considered in respect to a simple and effective data exploration task.

 

Biography:

Rachel Menzies is a lecturer and Head of Undergraduate Studies (Computing) at the School of Science and Engineering at the University of Dundee. Her research interests include user centred design with marginalised user groups, such as users with disabilities, as well as exploring novel interfaces, data visualisation and CS education. Rachel is an Accessibility and Usability Consultant with the Human Centred Computing Consultancy, run by the University of Dundee, and has worked for many large international clients as well as providing bespoke training sessions to small companies.

Event details

  • When: 23rd February 2016 14:00 - 15:00
  • Where: Cole 1.33
  • Series: School Seminar Series
  • Format: Seminar, Talk

The scientific life of Ada Lovelace

Prof. Ursula Martin will be talking about the letters of Lady Ada Lovelace at 5.30pm on Thursday 17th December in Abertay University (Kydd Building, Bell St, Dundee, behind Dundee High School).

This is a BCS sponsored event and all are welcome. Teas/Coffees from 4.30pm. PhD research posters will also be on display.

Event details

  • When: 17th December 2015 16:30 - 18:30
  • Format: Talk

Seminar: “Interaction, Embodiment and Technologies in Early Learning” by Dr Andrew Manches

as

Abstract: 

 

Most of us might agree that ‘hands-on learning’ is good for children in the early years. But why? Is it simply more fun and sociable, or are there any more direct cognitive benefits? And what determines definitions of ‘hands-on’? Can we include iPads? This talk will draw upon an ESRC-funded project to examine the educational implications of recent theoretical arguments about the embodied nature of cognition. Video data from the project will be used to illustrate the methodological significance of the way children gesture when describing mathematical concepts and evaluate a hypothesis that numerical development is grounded upon two particular embodied metaphors. If correct, this presents a serious challenge to traditional approaches to the types of learning materials we offer children. The talk then demonstrates two embodied technologies to consider the potential of new forms of digital interaction to further our understanding of embodied cognition as well as support early learning.

 

 Bio:  

 

Dr Andrew Manches is a Chancellor’s Fellow in the School of Education and leads the Children and Technology group at the University of Edinburgh.  He has 20 years experience working with children, first as a teacher, then as an academic. His recent research, funded by an ESRC Future Research Leader grant, focuses on the role of interaction in thinking, and the implications this has for early learning and new forms of technology.  When not being an academic, Andrew is a parent of two young children and directs an early learning technology start-up that was awarded a SMART grant this year to build an early years maths tangible technology.

Event details

  • When: 24th November 2015 14:00 - 15:00
  • Where: Cole 1.33a
  • Series: School Seminar Series
  • Format: Seminar, Talk

IEEE talks: JEEVES and the Speculative W@nderverse

In October 2015 we attended two IEEE conferences in the USA. Daniel Rough presented a full paper at the IEEE Symposium on Visual Languages and Human-Centric Computing, October 18–22, in Atlanta, Georgia. Uta Hinrichs presented a full paper at the IEEE Information Visualization (InfoVis) conference as part of VizWeek, October 25-30 in Chicago, Illinois. You can find details of these and all our papers on the publications page.

On November 10th, Daniel and Uta will reprise their talks here in St Andrews and everyone is welcome to attend. You can find the details for the two talks below along with links to the papers via the University of St Andrews research portal.

Talk 1 by Daniel Rough
Title: Jeeves – A Visual Programming Environment for Mobile Experience Sampling
Authors: Rough, D and Quigley, A.

Abstract: The Experience Sampling Method (ESM) captures participants’ thoughts and feelings in their everyday environments. Mobile and wearable technologies afford us opportunities to reach people using ESM in varying contexts. However, a lack of programming knowledge often hinders researchers in creating ESM applications. In practice, they rely on specialised tools for app creation. Our initial review of these tools indicates that most are expensive commercial services, and none utilise the full potential of sensors for creating context-aware applications.

We present “Jeeves”, a visual language to facilitate ESM application creation. Inspired by successful visual languages in literature, our block-based notation enables researchers to visually construct ESM study specifications. We demonstrate its applicability by replicating existing ESM studies found in medical and psychology literature. Our preliminary study with 20 participants demonstrates that both non-programmers and programmers are able to successfully utilise Jeeves. We discuss future work in extending Jeeves with alternative mobile technologies.

Paper details from the St Andrews Research Portal: Rough, DJ & Quigley, AJ 2015, ‘ Jeeves – a visual programming environment for mobile experience sampling ‘, IEEE Symposium on Visual Languages and Human-Centric Computing (VL/HCC), Atlanta, United States

 

Talk 2 by Uta Hinrichs

Title: Speculative Practices: Utilizing InfoVis to Explore Untapped Literary Collections

Authors:
Hinrichs, U.,  Forlini, S. ; Moynihan, B.
Abstract:
In this paper we exemplify how information visualization supports speculative thinking, hypotheses testing, and preliminary interpretation processes as part of literary research. While InfoVis has become a buzz topic in the digital humanities, skepticism remains about how effectively it integrates into and expands on traditional humanities research approaches. From an InfoVis perspective, we lack case studies that show the specific design challenges that make literary studies and humanities research at large a unique application area for information visualization. We examine these questions through our case study of the Speculative W@nderverse, a visualization tool that was designed to enable the analysis and exploration of an untapped literary collection consisting of thousands of science fiction short stories. We present the results of two empirical studies that involved general-interest readers and literary scholars who used the evolving visualization prototype as part of their research for over a year. Our findings suggest a design space for visualizing literary collections that is defined by (1) their academic and public relevance, (2) the tension between qualitative vs. quantitative methods of interpretation, (3) result- vs. process-driven approaches to InfoVis, and (4) the unique material and visual qualities of cultural collections. Through the Speculative W@nderverse we demonstrate how visualization can bridge these sometimes contradictory perspectives by cultivating curiosity and providing entry points into literary collections while, at the same time, supporting multiple aspects of humanities research processes.

Published in:
IEEE Transactions on Visualization and Computer Graphics,  (Vol:22, Issue: 1 )  Page(s): 429 – 438
ISSN :  1077-2626
DOI:     10.1109/TVCG.2015.2467452

Paper details from the St Andrews Research Portal:
Hinrichs, U, Forlini, S & Moynihan, B 2015, ‘ Speculative practices: utilizing InfoVis to explore untapped literary collections IEEE Transactions on Visualization and Computer Graphics .

Event details

  • When: 10th November 2015 14:00 - 15:00
  • Where: Cole 1.33a
  • Format: Talk

Dr Roy Dyckhoff, Hon. Senior Lecturer: ‘Coherentisation of first-order logic’

The School of Computer Science is delighted to announce that honorable lecturer Dr Roy Dyckhoff is an invited speaker at the conference on Automated Reasoning with Analytic Tableaux and Related Methods, in Wrocław (Poland) from 20–24 September.

roy

Abstract: This talk explores the relationship between coherent (aka “geometric”) logic and first-order logic (FOL), with special reference to the coherence/geometricity required of accessibility conditions in Negri’s work on modal logic (and our joint work with her on intermediate logic). It has been known to some since the 1970s that every first-order theory has a coherent conservative extension, and weaker versions of this result have been used in association with the automation of coherent logic; but, it is hard to find the result in the literature. We discuss various proofs of the result, and present a coherentisation algorithm with the desirable property of being idempotent.

Please see: http://tableaux2015.ii.uni.wroc.pl/index.html for further details

Event details

  • When: 20th September 2015 10:00 - 24th September 2015 17:00
  • Format: Conference, Talk

Seminar: ‘Formalizing Garbage: Mathematical Models of Memory Management’ by Jeremy Singer

Abstract:

Garbage collection is no longer an esoteric research interest. Mainstream programming languages like Java and C# rely on high-performance memory managed run time systems. In this talk, I will motivate the need for rigorous models of memory management to enable more powerful analysis and optimization techniques. I will draw on a diverse range of topics including thermodynamics, economics, machine learning and control theory.

Bio:

Jeremy Singer is a lecturer at the School of Computing Science, University of Glasgow, Scotland. He has research interests in programming languages,compilation, run time code optimization and memory management. Singer received his PhD from Cambridge in 2006. Website:http://www.dcs.gla.ac.uk/~jsinger

 

Event details

  • When: 6th October 2015 14:00 - 15:00
  • Where: Cole 1.33a
  • Series: School Seminar Series
  • Format: Seminar, Talk

Talk: ‘This is a Google Talk’ – by James Smith

The School of Computer Science is pleased to welcome back one of its former PhD students, James Smith, who is currently Google Product Manager, London to talk about Google @ St Andrews. FB_20150916_14_52_25_Saved_Picture

Abstract: Google’s mission is to organize the world’s information and make it universally accessible and useful. It’s an enormous goal to accomplish and we need great people to help us achieve it. We invite you to come learn about Google and some of the engineering challenges we’re tackling.

Sign up: goo.gl/GXXiWp

Event details

  • When: 13th October 2015 20:00 - 21:00
  • Where: Cole 1.33
  • Format: Talk

School Seminars: Building the News Search Engine – Bloomberg

Building the news search engine, by Ramkumar Aiyengar, Bloomberg
Abstract:
This talk provides an insight into the challenges involved in providing near real-time news search to Bloomberg customers. Our News team is in the process of migrating to using Solr/Lucene as its search and alerting backend. This talk starts with a picture of what’s involved in building such a backend, then delves into what makes up a search engine, and then discusses the challenges of scaling up for low-latency and high-load.
Bio:
Ramkumar leads the News Search backend team at the Bloomberg R&D office in London. He joined Bloomberg from his university in India and has been with the News R&D team for 7 years now. For the last couple of years, his team has focussed on rewriting almost the entire search/alert backend, used by almost every Bloomberg user to get near-real time access to news with sub-second latencies. A geek at heart, he considers himself a Linux evangelist, an open source enthusiast, and one of those weird creatures who believes that Emacs is an operating system and had once got his music player and playlists to be controlled through a library written in Lisp.

Event details

  • When: 3rd March 2015 14:00 - 15:00
  • Where: Cole 1.33
  • Series: CS Colloquia Series, School Seminar Series
  • Format: Seminar, Talk

School Seminar Series: Matching in Practice: Junior Doctor Allocation and Kidney Exchange

Matching in Practice: Junior Doctor Allocation and Kidney Exchange by Dr. David Manlove

Abstract:
Matching problems typically involve assigning agents to commodities, possibly on the basis of ordinal preferences or other metrics. These problems have large-scale applications to centralised matching schemes in many countries and contexts. In this talk I will describe the matching problems featuring in two such schemes in the UK that have involved collaborations between the National Health Service and the University of Glasgow. One of these dealt with the allocation of junior doctors to Scottish hospitals (1999-2012), and the other is concerned with finding kidney exchanges among incompatible donor-patient pairs across the UK (2007-date). In each case I will describe the applications, present the underlying algorithmic problems, outline the computational methods for their solution and give an overview of results arising from real data connected with the matching schemes in recent years.

BIO:
David Manlove is a Senior Lecturer at the School of Computing Science, University of Glasgow, where he has been since 1995. His research interests lie mainly in the field of algorithms and complexity, and include algorithms for matching problems involving preferences. These arise in applications such as the assignment of school leavers to universities, kidney patients to donors and junior doctors to hospitals. He and his colleagues have been involved in collaborations with the NHS in relation to the Scottish Foundation Allocation Scheme (for matching junior doctors to hospitals) and the National Living Donor Kidney Sharing Schemes (for enabling kidney “swaps” between incompatible donor-patient pairs) where optimal matching algorithms developed by him and colleagues have been deployed. He has over 50 publications in this area including his book “Algorithmics of Matching Under Preferences”, published in 2013.

Event details

  • When: 3rd February 2015 14:00 - 15:00
  • Where: Cole 1.33a
  • Series: School Seminar Series
  • Format: Talk