Department of Computing Seminars 2000/2001
2000
Checking States and Transitions of a set of Communicating Finite State Machines
Tuesday 24th October, 4.30pm
Dr Robert M. Hierons, Department of Information Systems and Computing, Brunel University.
Augmenting Connectivity in Graphs
Tuesday 7th November, 4.30pm
Dr Tibor Jordan, Department of Operations Research, Etovos University, Budapest, Hungary.
Channel assignment and multicolouring of subgraphs of the triangular lattice
Tuesday 14th November, 4.30pm
Frederic Havet,Oxford
Set functions and clustering: quasi-convexity and greedy optimization with monotone linkage functions.
Tuesday 28th November, 4.30pm
Prof. Boris Mirkin, Birkbeck College, University of London
In data of complex structure, such as graph or image or protein fold, a
linkage function d(i,S) may capture important information on similarity
between elements i and subsets S of a finite set I. A set function
F(S), defined as minimum of d(i,S) over all i from S (or, in a
different setting, over all i from I-S), satisfies a so-called
quasi-convexity condition if and only if d(i,S) is monotone over S.
Greedily seriating I according to monotone d(i,S) leads to global
maxima of F(S), which can be used for finding cluster patterns in data.
Secret sharing with disenrolment.
Tuesday 5th December, 4.30pm
Prof. Peter Wild, Department of Mathematics, Royal Holloway, University of London.
The NewsTools Project
Tuesday 12th December, 4.30pm
Dr Anthony Hunter, Dept of Computer Science, UCL, University of London.
The NewsTools Project is aimed at developing tools for analysing and
merging news reports that are represented in the form of structured
text.
Structured text is an idea implicit in a number of approaches to
handling information such as news reports. An item of structured text
is a set of semantic labels together with a word, phrase, sentence,
null value, or a nested item of structured text, associated with each
semantic label. As a simple example, a report on a corporate
acquisition could use semantic labels such as "buyer", "seller",
"acquisition", "value", and "date".
Some news agencies store news reports as structured text. In addition,
new technologies, such as information extraction and XML, will
massively increase the amount of structured text available. The advent
of structured text raises the need for tools for structured text
analysis. This project focuses on merging news reports and on analysis
for identifying and acting on incompleteness and inconsistency in news
reports.
The tools being developed in the NewsTools Project are for the following tasks:
- Merging multiple news reports
- from heterogeneous sources
- collected over time
- Handling inconsistency in news reports
- identifying inconsistency
- measuring the degree of inconsistency
- locating the source of inconsistency
- acting on inconsistency
- Handling incompleteness in news reports
- Reasoning with news reports
- identifying ramifications of news reports
- constructing arguments using news reports
These
tools are being developed using a number of ideas from artificial
intelligence, computational linguistics, and data and knowledge
engineering. In particular, they are being developed using logic-based
knowledge representation and reasoning techniques. These techniques
allow for a formal specification of the functionality of the tools, and
also support a viable approach to managing the necessary domain
knowledge required for the tool functionality.
The domain knowledge
is specialized knowledge about the information in news reports. Given
the complexity in acquiring and maintaining this kind of knowledge, it
is only possible to capture knowledge on very focussed domains. We are
currently working on certain kinds of business news reports, including
mergers and acquisitions news, where there are a number of regularities
in the format of the news reports and of the kinds of domain knowledge
required.
The coverage of the domain knowledge depends on the application area
but includes context-sensitive lexical knowledge (eg. synonymous words
and phrases, more general terms, more specialized terms, preferred
terms, meronyms, etc) and context-sensitive world knowledge (eg.
concept definitions, relationships between concepts, geographic
knowledge, knowledge about calendars and time, knowledge about people
and organizations, commonsense knowledge, etc).
Currently the focus of the project is in on developing the theoretical
foundations of the NewsTools: Developing an architecture for the tools
and the associated domain knowledge; Acquiring test sets of structured
news reports; and Compiling lexical and world knowledgebases.
The theoretical work has benefited from EPSRC and ESPRIT funding. The
work has also benefited from collaborative links with companies
including Reuters, Factiva, TFPL and Railtrack. Currently, funding is
being sought to start implementing prototype tools and undertake
evaluative studies.
For more information about the NewsTools Project, contact Tony Hunter (a.hunter@cs.ucl.ac.uk or +44 20 7380 7295).
Tuesday 30th January, 4.30pm
Dr Adam Kilgarriff, University of Brighton.
Semigroups.
Tuesday 20th Feburary, 4.30pm
Prof. Scerbakov, Visiting Department of Mathematics, University of Reading.
2001
Language is Never, Ever, Ever RandomTuesday 30th January, 4.30pm
Dr Adam Kilgarriff, University of Brighton.
An Introduction to the Theory of Quasigroups, and Applications.
Tuesday 20th Feburary, 4.30pm
Prof. Scerbakov, Visiting Department of Mathematics, University of Reading.
Some connections between Bernoulli convolutions and analytic number theory.
Tuesday 27 Feburary, 4.30 pm
Dr T. Hilberdink, University of Reading
Program Schemes, Finite Model Theory and Computational Complexity
Tuesday 6th March, 4.30 pm
Prof Iain Stewart, University of Leicester .
Analyzing the cache behaviour of non-uniform distribution sorting algorithms
Tuesday 13th March, 4.45 pm
Professor Rajeev Raman, Computer Science Group, Dept of Maths & Comp. Sci, University of Leicester, Leicester, LE1 7RH.
Extensions of the Stone-Weierstrass Theorem
Tuesday 20th March
Speaker: Dr Barnaby Sheppard (University of North London)
Experimental evaluation of algorithmic solutions for genrealized network flow problems.
Tuesday 15th May
Dr Tomasz Radzik, Dept of Computer Science, Kings College.
Fragmentability of graphs
Tuesday 22nd May
Prof. Graham Farr, School of Computer Science, Monash University, Victoria, Australia