[Ltg] LTG Seminar [Jette Viethen, 11-12-2006, E6A 357,11am]

[Marc Tilbrook]

----
 LTG Seminar
  - see: http://www.clt.mq.edu.au/Events/Seminars.html

  Monday, 11th December, 2006, 11am
  Macquarie University, E6A, Room 357
 ----

-----
Note: This is the last seminar for 2006. We start 2007 with Martin Volk on 
15th January.
-----

Title: Evaluation, Graphs and Relations in Referring Expression Generation
Speaker: Jette Viethen

Based on the experiment I presented in my last LTG talk (dry-run for
INLG in July), I will give an overview of issues that arise for
evaluation of referring expression generation systems and more
generally Natural Language Generation systems. The second section of
my talk will cover work very much in progress on making use of the
graph-based framework for referring expression generation by Krahmer
et al. (2003). I am especially interested in using this framework to
generate descriptions that involve relations between objects to
overcome shortcomings of Dale and Haddock's (1991) relational
algorithm. If time permits, I would like to give an overview over
further work I have planned for my PhD project to get feedback from
the group.


----
Before Jette's talk, Anders Tychsen will do a brief 5 minute presentation
----

Games and Communication.
As most of you will know, either from being directly involved in the
experimental work or via the department grapewine, I am currently
working with multi-player role playing games. As part of this project, I
have recorded a series of game sessions, and from these recordings
produced series of transcripts of the conversation and game actions of
the players. By talking to some of you in the LT group, it has become
apparent that the kind of data I am working with might be of interest in
a LT context, i.e. the application of sentiment analysis to transcripts
in order to examine how the use of language varies between different
game segments. Although this type of analysis does not form a key
component of the project, I am very interested in learning about
possible LT analyses or techniques that can be applied to these forms of
data, the research questions that they can address, and perhaps form a
collaboration on some of these subjects. The point of the very brief
presentation is therefore to provide a quick overview of the current
project, and ask for your ideas, input, suggestions etc.