CALL FOR PAPERS
SPECIAL ISSUE
of
COMPUTATIONAL LINGUISTICS
on
QUESTION ANSWERING IN RESTRICTED DOMAINS
GUEST EDITORS
* Diego Mollá
Macquarie University
diego@ics.mq.edu.au
* José Luis Vicedo
Alicante University
vicedo@dlsi.ua.es
TOPIC AREA
In early descriptions of AI problems, question answering (QA) was
typically used to illustrate Natural Language Understanding (NLU)
tasks. It could be argued that QA is an ideal task to advance
knowledge in inference, NLU, and Computational Linguistics (CL) in
general.
There has been a recent surge of interest in research in QA, but much
of that research focuses on the mining of answers from open-domain
text collections. A consequence of this focused research is the
development of redundancy-based techniques that take advantage of the
enormous amount of information found in large corpora. Some concerns
have been raised as to whether the use of large corpora and generic
open-domain document sets is appropriate as a way to advance research
in natural language processing (NLP).
The use of restricted domains, on the other hand, presents interesting
challenges and opportunities that may take research to a new stage.
A clear challenge to the use of restricted domains (e.g. law,
medicine, technical manuals) is the diversity of these
domains. Different domains may present different stylistic
conventions. Also, restricted domains may use terminology that is not
stored in conventional lexical resources. As a consequence, approaches
devised for open-domain systems may encounter difficulties when
applied to these specific domains, thus raising the question of how
portable and re-usable these systems can be, and, on the other hand,
which kinds of additional or new NLP techniques are needed.
The most salient opportunities derive from the nature of the
restricted domains and the sorts of questions that are asked in these
domains. Restricted domains enable the development and use of
knowledge and lexical resources that would be impossible to produce
for open domains. Moreover, the kind of questions users desire to pose
to the QA system are dependent on the domain, and typically they
require a more complex processing than the "factual" questions
generally used in the common evaluations of open-domain QA. Restricted
domains are therefore ideal for the development of logic-based
approaches and the integration of reasoning methods that would handle
questions requiring complex inferences.
TOPICS OF INTEREST
-Comparisons between open-domain and restricted-domain QA.
-Characterisations of types of domains and technology required for QA
on those domains.
-Portability of QA systems between different domains.
-Generation of answers from multiple documents.
-Use of ontologies.
-Inference and reasoning.
-Question and information source analysis and representation.
-Knowledge representation.
-Answer validation.
-Question type classification and analysis.
-Evaluation.
Papers should not simply describe an existing system. Of primary
interest is the theoretical basis of the work presented. We will
especially welcome papers that show the impact of the above topics on
aspects of QA in restricted domains that may give an insight towards
advanced research in CL and NLP.
SCHEDULE
31 Jan 2005 - Call for papers issued
4 Jul 2005 - Papers due
17 Oct 2005 - Notification to authors
SUBMISSION PROCESS
Only electronic submission will be accepted. All submissions should be
sent to the CL journal <compling@ics.mq.edu.au> in accordance with the
instructions provided at http://www.aclweb.org/cl/; in the subject
line of your email, please ensure that you indicate that the paper is
intended for the QA Special Issue.
In addition to following the procedure described on the web site,
authors should also send the abstract of their paper electronically to
the two guest editors: <diego@ics.mq.edu.au>, <vicedo@dlsi.ua.es>.
Questions about the submission process should be addressed to
diego@ics.mq.edu.au.
Each submitted paper will be reviewed by two reviewers appointed by
the editor of CL and by two reviewers selected by the guest editors.
This archive was generated by hypermail 2b29 : Mon Jan 31 2005 - 09:34:10 MET