[Ltg] LTG Seminar [Mark Dras, 25-09-2006, E6A 357, 11am]

[James Ballantine]

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

  Monday, 25th September, 2006, 11am
  Macquarie University, E6A, Room 357
----

Title: Synchronous TAG -- Some Formal Stuff

Speaker: Mark Dras

When you're mapping between structures -- e.g. for translation or 
paraphrase -- there are a lot of different ways you could define the 
mappings, which will have implications for computational and mathematical 
properties such as complexity and formalism generative capacity. 
Synchronous Tree Adjoining Grammar (S-TAG) defines one particular way of 
mapping between TAG trees, with the aim being for the formalism to have 
some nice mathematical properties.  That the properties do hold is a 
consequence of some work from earlier this year by Stuart Shieber at EACL 
using a newly defined bimorphic tree transducer; what I'm interested is 
using some classical tree automaton results to do a similar thing, with 
the idea that this can be extended to some more general mappings.

So it's work in progress.  I'll explain TAG and S-TAG, what the general 
problem is, what tree automaton tools I'll be working with and the 
modifications to them, and the general outline of the approach I'll be 
taking.  It'll be lots of fun.