PhD defence in automatic SL processing

[Michael Filhol]

Dear all,

I am happy to inform you that I will defending my PhD on Friday, June 
27, at Limsi in Orsay, France. You are invited in building 508 of the 
Paris-11 University campus. Below is a summary of the thesis.

Thank you, and see some of you there maybe.

--
Formal approaches to representing the signs of Sign Languages are 
traditionally parametric and this work shows that they are
inappropriate for use in computer science. The main reasons are: the 
parameters used are neither all necessary nor do they form a sufficient 
set; parameters take on fixed values whereas signs are dynamic in nature 
and values change through time; parametric descriptions do not account 
for the signs' adaptability to context, hence are not reusable, which 
brings them to disregard the power in concision of sign languages. We 
propose a model called Zebedee, which describes signs in a sequence of 
timing units, each of which specifies a set of necessary and sufficient 
constraints to apply to a skeleton. The signing space is regarded as a 
Euclidean geometric space where any auxiliary geometric object may be 
built. Dependencies between elements of the descriptions or indeed on 
context are not only possible but also made relevant, and are based on 
articulatory, semantic and cognitive issues. We then give two 
complementary processes for evaluation: in computer science with the 
implementation of Zebedee in a signing avatar animation platform and an 
information display system for train stations, and in linguistics with a 
data base and new possibilities of queries that linguists may want to 
test. As prospects, we discuss different computational fields in which 
Zebedee should be useful, and several present linguistic problems for 
which it holds pieces of solutions.

-- 
Michael Filhol
_______________________________________________
SLLING-L mailing list
SLLING-L at majordomo.valenciacc.edu
http://majordomo.valenciacc.edu/mailman/listinfo/slling-l