The Flatlands meeting is an annual meeting of the NLP groups at Cambridge, Essex, Open and Oxford universities. It is an opportunity for these communities to meet and learn about recent work by the research students at the four sites (through presentations).
Previous meetings have been at Cambridge (2005), Oxford (2006) and the Open University's London offices (2007).
This year's Flatlands meeting will take place at the University of Essex on the 6th of June.
The programme, with links to most of the presentations, can be found below, where you will also find abstracts for most of the talks.
The organizers at Essex are: Udo Kruschwitz, Massimo Poesio & Doug Arnold.
Please contact the organisers at if you have any queries.
Date: 6th of June, 2008.
Location:
The meeting will be held at the University of Essex (on the Colchester Campus), in the Seminar Room of the Networks Building (room 1N1.4.1).
For information on getting to the University (Colchester Campus), see http://www.essex.ac.uk/visiting/colchester.aspx.
To help you find the Networks building, there is a Campus map (Networks building in bottom left hand corner), at http://www.essex.ac.uk/colchester/guide.aspx.
(The titles of talks are linked to the accompanying slides, where they are available; otherwise, the abstracts for most of the talks can be found below).
Dyaa Albakour (Essex): Intelligent Mail Server for Cross-organization collaboration.
Richard Bergmair (Cambridge): Monte Carlo Semantics: Robust Inference and Logical Pattern Processing Based on Integrated Deep and Shallow Semantics.
Abstract:
In this talk, I will be presenting some work currently in progress
as part of my PhD project. Its aim is to develop both a theory and
a practical and effective approach to robust inference and logical
pattern processing on textual data.
The work is situated in the broader research context of computational semantics and is centered around the (R)MRS semantic representation language for integrated deep and shallow processing.
This talk, in particular, will focus on the theoretic side of my work, describing a logic, an epistemology, and an algorithm for determining the logical/semantic degree of similarity or entailment for two given (R)MRSs.
Eva Banik (OU): An Integrated Architecture for Generating Parenthetical Constructions.
Jon Chamberlain (Essex): Creating annotated resources through Web collaboration.
Kakia Chatsiou (Essex): GreekGram: reporting on the progress of the implementation of a fragment of Modern Greek Grammar using the XLE parser.
Abstract:
GreekGram is a preliminary effort to implement a computational grammar for Modern Greek, (currently in a fragmentary stage) following the principles of the Lexical Functional Grammar (LFG) Parallel Grammar (ParGram) Project [1], a collaborative effort among researchers in industrial and academic institutions around the world with the aim of producing wide coverage grammars for a wide variety of languages. The long-term goal of the project is a compilation of a wide coverage Lexical Functional Grammar (LFG) of Modern Greek using the Xerox Linguistics Environment (XLE) parser [2] and at its current stage of development covers the syntax of basic word order phenomena in Modern Greek, relative clauses and coordination. In our presentation, we will present an overview of the GreekGram Project, review the methods and tools employed as well as some of the assumptions underlying the current fragment. Finally, we will demo the current stable version of the fragment and suggest possible future development directions and practical application in the areas of machine translation, and Natural Language Processing industry.
References:
[1] The Parallel Grammar Project. http://www2.parc.com/isl/groups/nltt/pargram/
[2] The XLE parser. http://www2.parc.com/isl/groups/nltt/xle/
Mohan Ganesalingam (Cambridge): Linguistic phenomena in mathematics.
Abstract:
This work is part of a long-term project (joint work with Thomas
Barnet-Lamb) to create a fully formal language for representing
mathematics. I will begin by showing previous attempts at such
languages and demonstrating that one can dramatically improve on these
by fusing concepts from formal languages and natural languages to
create a new kind of language: a natural formal language.
The body of the talk will then discuss a single example, namely quantification of variables. I will present an argument which demonstrates that the mechanisms used to handle quantification in formal languages cannot cope with quantification in mathematics; I will then show how a theory of natural language semantics, Discourse Representation Theory, can be combined with formal concepts to generate the correct predictions.
Brian Harrington (Oxford): ASKNet: Creating and Evaluating Large Scale Integrated Semantic Networks.
Deirdre Lungley (Essex): Adaptively Modelling the Context of an Intranet Query.
Yue Zhang (Oxford): Joint word segmentation and POS-tagging on a single perceptron.
Abstract:
For Chinese POS tagging, word segmentation is a preliminary step. To avoid
error propagation and improve segmentation by utilizing POS information,
segmentation and tagging can be performed simultaneously. A challenge for this
joint approach is the large combined search space, which makes efficient
decoding very hard. Recent research has explored the integration of
segmentation and POS tagging, by decoding under restricted versions of the full
combined search space. In this paper, we propose a joint segmentation and POS
tagging model that does not impose any hard constraints on the interaction
between word and POS information. Fast decoding is achieved by using a novel
multiple-beam search algorithm. The system uses a discriminative statistical
model, trained using the generalized perceptron algorithm. The joint model
gives an error reduction in segmentation accuracy of 14.6% and an error
reduction in tagging accuracy of 12.2%, compared to the traditional pipeline
approach.