Pieter Adriaans | | Grammar induction and adaptive information disclosure
The adaptive information disclosure (AID) project is part of a larger
effort that aims at the creation of a so-called Virtual Lab environment
for e-science (VL-E). In the context of AID we are building in the
coming 4 years a suite of dynamic model driven information and knowledge
extraction tools on top of an architecture for grid-based distributed
data analysis. Keywords are: semantic models, agent technology, formal
concept analysis, datamining, textmining, gridmining, grammar induction,
question answering and the dynamic maintenance of ontologies. In my talk
I will focus on own contribution regarding the integration of grammar
induction and semantic learning in this context. Key research challenges
are:
-
How to organize the content of domain specific knowledge: The creation
and maintenance of semantics models of complex scientific domains on
the basis of a multitude of heterogeneous sources of information.
-
How to deploy the models in the knowledge extraction process: The
availability of perfect models of the domain does not guarantee a
performance boost of existing knowledge extraction algorithms. For
each algorithm a careful analysis of the function of the model
information in various stages of its execution has to be
made. Theoretical assumptions about complexity issues and search bias
have to be validated against empirical evidence.
-
Integration with Grid technologies: The possibility of implementing
this process on the basis of an architecture of co-operative agents in
a grid.
|