proposing a AAAI Spring Symposium 2007 on Cognitive Approaches to NLP ...

[Krishna Jha]

A symposium topic entitled "Cognitive Approaches to NLP" has been submitted
to the AAAI Spring Symposium 2007 selection committee - full details below.
This symposium is a follow up to the very successful AAAI 2006 Spring Symposium
"Between a Rock and a Hard Place: Cognitive Science Principles Meet AI Hard Problems"
chaired by Christian Lebiere and Bob Wray. We are currently looking for
people who might be interested in participating in this symposium (no
commitment to attend is required at this point). Please let us know if you
would like to be listed as a potential participant, by contacting Jerry Ball ( jerry.ball at mesa.afmc.af.mil ) or Krishna Jha ( kjha at atl.lmco.com ) by 28 April 2006.

Thank you,
Jerry Ball, Air Force Research Laboratory
Krishna Jha, Lockheed Martin Advanced Technologies Laboratories

ps: Our apologies in advance for any overlapping postings.

------------------------
Full proposal:

Cognitive Approaches to NLP

As a follow up to the very successful AAAI 2006 Spring Symposium – Between a Rock and a Hard Place: Cognitive Science Principles Meet AI Hard Problems (Lebiere & Wray, 2006) – we are planning a more specialized symposium with the title – Cognitive Approaches to NLP.

Current symbolic approaches to NLP continue to suffer from problems of brittleness and unscalability and statistical/connectionist approaches have so far failed to adequately capture higher-level knowledge. Consequently, we have been unable to build practical NLP systems for large, real-life applications. However, there is an acute need for robust NLP systems that can be applied to large scale (e.g., web scale) multidomain documents for a variety of applications (e.g., intelligence gathering, customer service). Human-inspired cognitive approaches which evidence a hybrid of symbolic and statistical/connectionist representations combined with a mixture of serial and parallel processing mechanisms have appeared promising for a long time, but details of human language representation and processing have been lacking. Luckily, over the past few years (and decades), research in cognitive science appears to have made enough progress so that we can now begin to apply and, perhaps, ev

aluate cognitive approaches towards building robust, large scale NLP systems. We propose this symposium to highlight NLP research at the intersection of AI/Computational Linguistics and Cognitive Science/Computational Psycholinguistics—especially research which integrates symbolic and statistical/connectionist representations with serial and parallel processing mechanisms into cognitively motivated NLP systems. Major topics of discussion are expected to include:
a)	Cognitive/hybrid approaches to semantic knowledge representation (KR)
b)	Cognitive/hybrid approaches to knowledge acquisition
c)	Cognitive/hybrid approaches to parsing
d)	Cognitive/hybrid approaches to generation
e)	Cognitively-motivated implementations of IE systems
f)	Cognitively-motivated implementations of NLP systems
g)	Applying cognitive architectures to build NLP systems
h)	Evaluations, advantages and disadvantages of cognitively motivated NLP systems

Research in the development of cognitively plausible NLP systems is necessarily interdisciplinary, drawing on theoretical advances in AI/computer science, linguistics, and psychology, among some of the more important disciplines. From an AI/computer science perspective, the interest is in overcoming the combinatorial explosion that is characteristic of solutions to hard problems like NLP via adoption of non-optimal solutions, appropriate representations (symbolic and/or statistical/connectionist), and heuristic techniques (including mixtures of serial and parallel processing mechanisms) capable of providing reasonable solutions in real-time given the computational and storage capacity of the brain—with human language capabilities providing the existence proof that such a system is possible. From a linguistic perspective, this research agenda emphasizes the relevance of linguistic performance and language processing to issues of linguistic competence. From the perspective of 

psychology, the research agenda acknowledges the contribution that computational implementations provide in focusing research on solving the most pressing problems.

This symposium will be of interest to researchers building cognitively motivated, large scale, real world NLP systems. Researchers engaged in this interdisciplinary area of research will be consumers of experimental research in psycholinguistics and human language processing without necessarily being psycholinguists or experimental psychologists. Their attempts to build computational implementations are likely to reveal weaknesses in cognitive theories of language representation and processing which will need to be investigated and further elaborated. This is especially likely to follow from attempts to build large-scale NLP systems—something that has not often been attempted in the cognitive science and experimental psychology communities. Issues of computational complexity, storage capacity and resource limitations which do not arise in the development of cognitive theories, perhaps implemented as small scale systems, are brought to the fore in the development of large sca

le systems.

Participants will be expected to present theoretical and applied research relevant to this symposium. Contributions can be in the form of either full-size 6-page papers or short 2-page white papers. The symposium will be structured into a number of sessions each centered around a common topic. Each session will include two to four presentations followed by brief comments from the session moderator and substantial discussion between presenters and symposium participants.


Organizers:

Jerry Ball
Senior Research Psychologist
Air Force Research Laboratory
6030 S. Kent St
Mesa, AZ 85212-6061
Jerry.Ball at mesa.afmc.af.mil

Krishna Jha
Lockheed Martin Advanced Technologies Laboratories
3 Executive Campus, 6th Floor
Cherry Hill, NJ 08002
kjha at atl.lmco.com

Sergei Nirenburg
Professor, CSEE
Director, Institute for Language and Information Technologies
University of Maryland Baltimore County
1000 Hilltop Circle
Baltimore, MD 21250
sergei at umbc.edu

Marjorie McShane
Department of Computer Science and Electrical Engineering 
ITE 325 
1000 Hilltop Circle 
University of Maryland, Baltimore County 
Baltimore, MD 21250
marge at umbc.edu

-- 
 Krishna Jha                    |	Email:	kjha at atl.lmco.com
 Lockheed Martin                |       Phone:	(856)792-9765
 Advanced Technology Labs       |       Fax:	(856)792-9925
 3 Executive Campus             |       
 Cherry Hill, NJ 08002          |