Dialogue and Natural Language Generation
Student Project Proposals
The term project/paper is an assessable component of the module,
carrying 50% of the total mark. Between 40-60 hours is allocated to
it. The topic for the project/paper should be chosen to reflect the
student's interests, for instance, it could reflect the student's
intended MSc or PhD topic.
Project Due Date: 8 April 2005
The term project/paper is designed to help you develop the
following skills:
- Project design. Cearly articulate the main aims of the
project. Break the project down into tasks, estimate how much effort
each task will require and partially order them. Construct a workplan
and identify milestones for yourself.
- Motivating research. Explain why this project is of
interest. Place it into the context of related previous and current
research. Although your term project will not be a complete end-to-end
system ready for use, explain how what you've done would fit into a
larger context. What applications could it serve?
- Well organised project report. Writing a succinct,
well-structured project report or critical review of the literature
is challenging. For a critical survey, you must avoid the
temptation to summarise everything you have read. Instead, you
should synthesise material from many sources to give an accurate
picture of the state of the art in the the topic area you have
chosen. Point out similarities among approaches, identifying
generalisations where possible. Critically compare and contrast
different approaches. Where possible, identify flaws in approaches
and describe how subsequent research was intended to overcome
limitations of prior approaches. Identify open problems in the
topic area. You will find this skill very valuable, e.g., for
writing conference and journal papers, MSc or PhD disserations, and
general report writing in whatever career you choose. If you choose
to do a programming project, you must resist the temptation to
simply document your program. Instead, motivate the work and
describe the algorithm(s) you developed and how you evaluated your
results.
- Analytical reading.
You should read each paper with a specific series of questions in
mind: what are the claims made for this work? what kind of claims
are they? what is the evidence for these claims? is the evidence
sufficient? This will help you analyse and critically assess the
paper.
- Differential reading Reading a paper from beginning to
end, devoting the same amount of time to each section is usually not
optimal. Identify the critical parts of the paper: where the claims
are made, where the evidence is presented, the key idea, etc, and read
these parts in depth and maybe several times. Multiple passes are
sometimes useful, e.g. to get the general gist of a complex argument
on the first pass and understand it in more detail on subsequent
passes. It may only be necessary to skim other parts, e.g. background
material with which you are already familiar, unless the presentation
of this material is an issue. You may want to place yourself mentally
in the situation of other readers with a different background to
yourself, in order to assess how well the presentation will work for
them.
- Managing your time. 40-60 hours is not a lot of time
and can easily be dissipated. You must structure your activities to
make efficient use of your time. Leave plenty of time for writing
up your report. Do not leave everything until the week before the
deadline.
Your proposal should be no more than 3 pages in length, and
should cover the following points:
- Background. This should include motivation for the
project, i.e., what problem it is addressing, and a brief summary
of related research on the topic. What's already been done in this
area? For term papers, this list should be 8-12 papers that you
believe characterise the state of the art on your chosen topic.
These are the papers that describe the approaches you will be
critically reviewing. For projects, at least 2 related papers
should be identified. In addition, you should describe the
techniques or tools that you intend to use, e.g., sentence
generators, statistical techniques, learning algorithms, corpora,
etc.
- Project Description. State the aims of the project and
outline how they are to be achieved. Describe your workplan,
including milestones and a timetable.
- Evaluation. For projects, describe how you will
evaluate your program.
Johanna
Moore