Personal tools
You are here: Home Talks & Speakers

Talks & Speakers

Slides from some of the talks are now available with more to follow


Speaker Title
Anne H Anderson
HCRC: Ahead of our time?  Crossing  boundaries of institution, discipline and sector: advantages and challenges In this talk I hope to mix a bit of nostalgia for the way we were back when HCRC was beginning with some reflections on the lessons learned from attempting to tackle research in this kind of interdisciplinary cross-institution way.  The talk will review the research  policy landscape then and now and the implications for how we do our  work.  The talk will also include some data from my studies of other interdisciplinary and collaborative university-industry projects. As we look forward to challenging times for universities and research funding I will attempt to see if the way we approached interdisciplinarity and impact in HCRC provides any models for the future. 
Bob Carpenter
Whence Linguistic Data?  Inferring Ground Truth   along with Annotator Accuracy, Bias and Variability
The empirical approach to linguistic theory involves collecting data and annotating it according to a coding standard.  The ability of multiple annotators to consistently annotate new data reflects the applicability of the theory.    In this talk, I'll introduce a generative probabilistic model of the annotation process for categorical data.  Given a collection of annotated data, we can infer the true labels of items, the prevalence of some phenomenon (e.g. a given intonation or syntactic alternation), the accuracy and category bias of each annotator, and the codability of the theory as measured by the mean accuracy and bias of annotators and their variability.  Hierarchical model extensions allow us to model item labelling difficulty and take into account annotator background and experience.  I'll demonstrate the efficacy of the approach using expert and non-expert pools of annotators for simple linguistic labelling tasks such as textual inference, morphological tagging, and named-entity extraction.  I'll discuss applications such as monitoring an annotation effort, selecting items with active learning, and generating a probabilistic gold standard for machine learning training and evaluation.  
Justine Cassell
Gossip and Small Talk: a Short Social History of Epistemics, and of Virtual Humans Buccleuch Place in the 1980s was a social place.  We post-graduate and post-doctoral students visited one another in our various alcoves and dens, carrying tea, biccies, and the latest gossip.  We shared stories about whose quals were apparently slated for immediate publication, who had gotten into yet another tussle with a department chair and, of course, who was sleeping with whom.  It is perhaps not surprising that I have spent much of my professional career trying to recreate that atmosphere of gossip, small talk, and cross-cultural communion. . . among virtual humans.  In this talk I will describe some anecdotes from the old days, and some of my newer research on imbuing virtual humans with social skills.
Simon Garrod
Reminiscing about HCRC and its achievements
Initially, I will reminisce about setting up HCRC and the first few years of the Centre. I will then concentrate on some of HCRC’s achievements and show how these have influenced my own work. I will concentrate especially on how HCRC highlighted the challenge of dialogue for the language sciences. This led eventually to my joint work on dialogue mechanisms with Martin Pickering (PhD & postdoctoral student at HCRC in the 90s). I will discuss this work both in relation to the theory of interactive alignment and more recent work on interweaving production and comprehension processes during dialogue.
Lauri Karttunen
Sources of Lexical Inferences with Demo
Much of the formal work in semantics has historically concentrated on logical connectives, quantifiers, modals and negation. Lexical semantics, the meaning of ordinary words, words that are not on your search engine’s stop list, has always been a less attractive field of study in formal semantics. The distinctions are fuzzy (cup vs. mug), ambiguity is pervasive (carry has 40 verb senses in WordNet). But there are areas of lexical semantics that have receive a lot of attention lately in the context of computing textual inferences and question answering.

In this talk I will first review some of the success stories of lexical semantics in a computational setting and continue with a discussion of unsolved issues and general problems.
Marc Moens
This time it's personal
I arrived in Buccleuch Place in 1982, and the next 20 years I worked in 6 or 7 different university departments without ever changing address. I then left Buccleuch Place to do something entirely different, but I still have some strong connections (as well as a key to the front door, actually).

In this talk I will reminisce about what it was like in those early days, what we achieved against the background of a fast-changing political climate, what those achievements look like from outside Buccleuch Place, and - if I figure this out in time - why I left when I did.

I was asked to leave time for questions at the end, at which point I will reminisce on demand.

I am no longer 24.
Johanna Moore
Welcome & Introduction

Fernando Pereira
Learning on the Web
It is commonplace to say that the Web has changed everything. Machine learning researchers often say that their projects and results respond to that change with better methods for finding and organizing Web information. However, not much of the theory or even the current practice, of machine learning take the Web seriously. We continue to devote much effort to refining supervised learning, but the Web reality is that labeled data is hard to obtain, while unlabeled data is inexhaustible. We cling to the assumption that events are drawn independently from a fixed distribution, while all the Web data generation processes drift rapidly and involve many hidden correlations. Many of our theory and algorithms assume data representations of fixed dimension, while in fact the dimensionality of data, for example the number of distinct words in text, grows as the Web grows. While there has been much work recently on learning with sparse representations, the actual patterns of sparsity in Web data rarely considered. Those patterns might be very relevant to the communication costs of distributed learning algorithms, which are necessary at Web scale, but little work has been done on this.

Nevertheless, practical machine learning is thriving on the Web. Statistical machine translation has developed non-parametric algorithms that learn how to translate by mining the ever-growing volume of source documents and their translations that are created on the Web. Unsupervised learning methods infer useful latent semantic structure from the statistics of term co-occurrences in Web documents. Image search achieves improved ranking by learning from user responses to search results. In all those cases, Web scale demanded distributed algorithms.

I will review some of those practical successes to try to convince you that they are not just engineering feats, but also rich sources of new fundamental questions that we should be investigating.
Barry Richards
How little I have learnt about search and why I care
One of the traditional aims of AI is to offer a model of human intelligence, one key aspect of the model being intelligent search. Defining intelligent search, and what algorithm paradigms exemplify it, is an outstanding challenge. Another is understanding and predicting the performance of particular search algorithms. How these challenges are addressed and resolved will determine what AI can contribute to the study of one aspect of human intelligence. Cognitive science appears to face the same challenge. I shall review what little insight I have derived from designing and implementing many search algorithms; and I shall indicate where more "light" might be helpful.
Jeremy Seligman
Channels and Minds
Channel theory is a branch of logic that studies information flow. In this setting channels can be understood in a very general way so that human beings can be seen as creating channels between reality and our minds. The channels that bodies create, via sense perception but also cognitive classification, are between external events in the world and internal events in the minds. One can add a Kantian twist to this story by seeing the a priori as the structure of our internal classifications (of experience). The a posteriori can then be seen as the structure of our classifications of external events. The thesis I will defend in this talk is that all a priori structure can be seen as the image of the a posteriori, i.e., information flow from external to internal. Because we are embodied, external events create a channel between our internal experiences, explaining the coordination between our different conceptions of the a priori.
Keith Stenning
Language evolution: enlarging the picture
As befits an HCRC birthday, this is a plea for interdisciplinarity: 
Specifically injecting some biology into discussions of language, and more generally cognitive, evolution.  The most challenging feature of the study of cognitive evolution is the 'phenotype description problem'.  With behaviour there is generally a plethora of possible descriptions, each with quite different implications for function and selection pressures. I will argue that the biologically insightful description of language is as discourse---the connected employment of language in the creation of context.  This description encourages us to see ancestral planning capacities as the precursor of language.

Biology now sees evolution as due to changes in the developmental processes of ancestral species (evolutionary developmental biology or 'EvoDevo'). A number of changes in developmental processes constituted {\em H. Sapiens}.  It is these biologically striking changes which provide the frame in which the changes in language and cognition must be interpreted. I will try.
Henry S Thompson  & Ellen Gurman Bard
A brief history of Buccleuch Place, in four parts
Epistemics gave birth to Cognitive Science, which hosted HCRC, which gave rise to . . . all kinds of things.  Not forgetting Glasgow and Durham.  With help from our friends, we'll try to reconstruct the origins of interdisciplinary collaboration and look briefly at the early years of Epistemics, before reviewing the birth of HCRC itself, focussing on why we won the competition.  That leads us naturally on to the HCRC Map Task -- its nature and its role in the life of the institution.  We'll finish with a few remarks about the present relevance of several particularly beneficial aspects of the IRC funding HCRC received.
Annie Zaenen
Getting From Text to Reasoning, a Case Study
An analysis of from-to complements with so-called verbs of motion

Joint work with Danny Bobrow, Cleo Condoravdi, Lucas Champollion and Raphael Hoffmann

We argue that there are two broad classes of uses of from-to modifiers in descriptions of situations in the physical domain: uncorrelated and correlated ones: in the uncorrelated ones, the from-to expressions are used to indicate the extent of a temporal or spatial interval, in the correlated ones, a functional relation between a spatial or temporal domain and a spatial or scalar range is described where the function requires a line-up between an initial and a final point in the domain and in the range of the function.

After describing these two uses of from-to modifiers, we discuss how a system might go about distinguishing these two types of readings and sub-cases of each.

Examples discussed include:
The road went from Palo Alto to Menlo Park.
Eric went from Palo Alto to Menlo Park.
The temperature went from 20 to 40 degrees.
The room went from 20 to 40 degrees.
The meeting went from 5 to 6 pm.


Document Actions