Information About Project Proposals
Your proposal should be approximately 3-5
pages in length, and should help me
understand both what you hope to accomplish, and how you plan to
accomplish it.
Please use the following structure, to help me evaluate whether I think your proposal is both interesting and feasible:
-
Title Page: Include a title, state which of the four types of
projects you will do, identify any partners, and provide a short abstract. (Only one proposal
per team is needed.)
-
Intellectual Merit: Clearly and specifically state the goals of
the project, then explain your project's significance by comparing
your research to the prior literature. Do not just list and summarize
previous work; instead compare the prior work to your
proposed research. This section should provide
sufficient background to motivate your work, and should answer the
following questions: What problem are you trying to solve? Why is
solving this problem important? What do you hope to learn by
conducting this project?
Although you don't yet know your results, you should know their format
(e.g., "Method A outperforms Method B under condition X").
-
Resources: Describe the resources (data, systems,
knowledge resources, etc.) that you will use and your plans for
acquiring or creating them.
-
Research Plan: Describe your methodology as well as your plans for
evaluation, as appropriate for the type of project.
Then define at a technical level the tasks that need to be done
as part of your project. Each task should be numbered and described in
detail. If there is more than one team member, your proposal should
state what each person will do.
For example, if an empirical project, what experiments will you conduct? What do you hope to learn from each
experiment? What evaluation metrics will you use? What are some
appropriate baselines and/or upperbounds for the evaluation? Etc.
-
Deliverables: Describe the deliverables that you will
produce. Learned features or a learned model? Quantitative and/or
qualitative evaluations? Implemented software? A coding manual and annotated corpus? Etc.
-
Schedule: Describe the time required to conduct each of the tasks outlined in
your Research Plan, and estimate completion dates for
each of your Deliverables.
-
References: The prior sections should include citations to relevant literature that you both have and plan to read, which
are summarized in your bibliography.
These should include at least one meaningful citation to a paper
from the class readings to ensure that you work on
something related to the course.
-
Appendix: Discuss any related work you are doing for another
class or research project, and contrast that work with what you plan
to do for your course project. It is fine to do something related to
an existing project, but your course project may not be work or
results you (or a partner) are already performing for another class or
research project. For example, if you are already performing an
annotation study, you could propose to develop an annotation scheme
for related information, and add your new types of annotations to your
existing data. Or, if you are already working on an NLP system, you
could propose to augment it with processing that you did not already
plan to add this semester.
Below are specific comments about specific types of projects which are in
addition to the specifications above:
- Project Option 2: Specify the input and output of the
system. Also, specify the type of strategy your system will use to
perform its task, and the types of knowledge your system will use to
perform it. If you are implementing an existing algorithm, include
comments about which aspects you may focus on more than others, or
other ways you may modify the algorithm or its use. If you are
running new experiments, please state your hypothesis or hypotheses
about what your experiments will show.
-
Project Option 3: Specify the application system, the
linguistic knowledge the system will use, how it will use it, and
your hypotheses about how the linguistic knowledge will benefit the
application system. Also, specify how the system will acquire the
linguistic knowledge: will manual annotations be given as input? If
so, which ones? If it will acquire the knowledge automatically,
please specify how.
-
Project Option 4: Specify as clearly as you can exactly what
type of thing you plan to annotate. You must have clear,
declarative, specific statements of what you will annotate.
Identify the corpus you will annotate. Specify how much data you
plan to annotate, and what work, if any, will be involved in
preparing the corpus for annotation. Give a few examples of
annotations of real naturally occuring sentences from the
corpus. You won't be sure of your categories yet, so just use your
current, initial ideas. It's fine (and expected) if your categories
and definitions change between the proposal and final draft.
Identify the tool you will use for annotation. Will you use an
annotation system like Gate? Configure a tool using a speech
processing system such as Waversurfer? Or, a formal, text-based
representation that your evaluation code will access using regular
expressions?