AAAI Spring Symposium 2003
March 24-26, 2003
Palo Alto, CA
Endorsed by Sigdial
Special Interest Group on
Dialogue of the Association for Computational Linguistics
Endorsed by
Siggen
Special
Interest Group on Generation of the Association for Computational
Linguistics
Last updated 14 Mar 2002 with schedule and list of
accepted papers.
Contents
- Submissions due
- Acceptances mailed
- Final papers due
- Registration deadline
- Symposium
|
- Friday, October 4, 2002
- mid-November
- mid-December
- Friday, Feb. 14, 2003
- Mon.-Wed., March 24-26, 2003
|
The symposium will be held at Stanford University, home of the
last several AAAI Spring Symposia. Palo Alto is easily reachable from
both the San Francisco and San Jose airports.
Monday, March 24
9:00 AM - 5:30 PM Symposia
sessions
6:00 PM - 7:00 PM Reception
Tuesday, March 25
9:00 AM - 5:30 PM Symposia
sessions
6:00 PM - 7:30 PM Plenary session
Wednesday, March 26
9:00 AM - 12:30 PM Symposia
sessions
Call for papers
All types of dialogue systems, including spoken, written,
GUI-based and multimodal, have become more prevalent in recent years.
Workshops in the artificial intelligence and computational linguistics
communities have focused on the building and evaluation of these systems
in a variety of domains. At the same time, other workshops have provided
a forum for natural language generation research, discussing the
generation of everything from noun phrases to longer monologues.
The goal of this symposium is to bring together people involved
with all types of dialogue systems and natural language generation
researchers to discuss current challenges and improve existing techniques
for meeting them. We expect builders of dialogue systems to come away
with insight into problems and solutions already discovered by the NLG
community, and builders of generation systems to develop a new
appreciation of issues that arise in dialogue systems.
Relevant areas include generating responses for tutoring and
interactive help systems; entertainment applications such as animated
intelligent agents and generation of dialogue in narrative; spoken or
written responses in information retrieval and transaction processing;
and even computer-computer dialogue intended to be seen by humans.
We expect participation from both workers in both dialogue systems
and natural language generation to discuss salient issues in both
communities, including:
- System design
-
- Which issues are common to the design of all types of
dialogue systems, and which are specific to specific modalities, such as
spoken or written?
- Could there be a general API defining interactions between
dialogue managers and NLG systems so that one could swap out different
components in a "plug and play" manner?
- What is the proper role of statistical vs. logic-based
representation in NLG-based dialogue systems?
- Initiative, repair and other dialogue issues
-
- How does determination of initiative, such as in
mixed-initiative systems, interact with the generation of responses?
- How can NLG and NLU cooperate in the generation of
repairs?
- Dialogue models, user models, etc.
-
- How much does NLG depend on a dialogue model for
constructing responses?
- What is the relationship between user model and dialogue
model in this type of generation?
- Relationship to natural language understanding
-
- To what extent does satisfactory NLG in dialogue systems
depend on the availability of good natural language understanding?
- To what extent can NLG and NLU share resources?
- Customization and resource sharing
-
- How much customization of responses is desirable for
different applications? How much is practical?
- To what extent can different applications share syntax and
semantics?
- Templates and prompts
-
- Under what conditions is template use the preferred form of
generation?
- Are system designers interested in more advanced NLG
architectures, or do simple canned prompts suffice?
- Speech synthesis
-
- How does generation interact with speech synthesis?
- What information about information structure, prosody,
etc. should be part of NLG, and how is it passed to synthesis?
- When is it important to have high quality pre-recorded
audio vs. low quality but more flexible synthesis?
- Relationship to graphics and other modalities
-
- How is language generation synchronized with generation of
non-language components, such as graphics?
- Dialogue in narrative
-
- What issues does generation of narrative (including
character dialogue) share with "real-world" applications?
- Does narrative-based dialogue differ fundamentally from
task-based dialogue? How?
- Corpus use
-
- What approaches to corpus collection and analysis are most
needed in the development of NLG-based dialogue systems?
- Evaluation
-
- How can we best evaluate these complex systems and their
components?
The symposium will consist mainly of panel discussions, group
sessions, and a poster session, but there will also be time for
researchers to present their work.
Potential participants may submit technical papers between 2 and
10 pages in length. Papers can be theoretical or applied, with technical
details or positions, completed or in progress. Short papers can be an
extended abstract, a position statement, or a description of a proposed
demo or poster. We are looking forward to diverse perspectives such as
spoken vs. text, generation vs. dialogue structure, and statistical vs.
knowledge-based. (Papers exclusively on speech synthesis are outside the
range of this symposium.)
All accepted papers will appear in the working notes, which will
be distributed prior to the symposium.
- Electronic submission is required. Papers may be submitted in
PostScript or PDF.
- Submissions should be emailed to
nlgdial@cs.niu.edu, on or before
October 4, 2002.
- The subject line of your email should include the word
"Submission" and the name of your file.
- To facilitate the assignment of reviewers, please include the
following in your email:
- Title of paper
- Name and coordinates of each author
- Contact person (if not the first author)
- Page count
- General subject area, e.g. from the list above
- A short summary (1-3 sentences)
- Include the file itself as an attachment. The file name should
include an indication of the first author's name and the file type, e.g.
turing.ps.
- If you are a student and wish to apply for student support,
please so indicate and we will send you the appropriate forms. Decisions
about student support are independent of decisions on paper
acceptance.
Final papers will use
standard
AAAI format. Submissions may use this format or any easy-to-read
format, e.g. single column, 12 point type, with 1-inch margins.
Please follow the instructions below to ensure that your paper
will print correctly the first time.
- Make sure that your file does not contain any non-printable
characters. In particular, check that there are no HP commands at the
beginning or end.
- Embed any special fonts. (If you are using an HP driver,
setting your driver options to "archive format" will accomplish both this
and the preceding goal.)
- Use margins suitable for US letter paper.
- Make sure that sure that your file does not specifically
request A4 paper. (This is especially a problem with certain TeX
drivers.)
- Make sure your diagrams are legible both in print and
online.
- Special note to Latex users: To make your file legible online,
be sure to include the following command:
\usepackage{times}
- Please consult the
summary
of Acrobat distiller settings and the
detailed
Acrobat distiller information on the AAAI web site.
- Embed any special fonts.
- Set the default page size to US letter paper (not A4), and use
margins suitable for US letter paper.
- Make sure your diagrams are legible both in print and
online.
- Special note to Latex users: To make your file legible online,
be sure to include the following command:
\usepackage{times}
Student support
Students are encouraged to submit to the
symposium. A limited amount of money is available for full-time graduate
students who are unable to obtain sufficient funding from local
sources.
Please send email to either of the co-chairs, Charles Callaway (
callaway@itc.it) or Reva Freedman
(freedman@cs.niu.edu).
Organizing committee