| |
|
|
The ideal proposal would:
- Be delivered as a .pdf
- List the group name and team members. It may seem stupid to name
your team, but 1) it's kinda fun and 2) we need some way to refer
to groups, so please come up with something.
- State clearly the problem you are addressing and what you hope to
create as an artifact. Technically, you could satisfy this requirement with
a single sentence (and some groups did with the preproposal), but this time
I'm looking for much greater detail. If your project has a big UI
component, can you include a figure showing what it might look like? Can
you include an architectural diagram? The more detail the better, and of
course you can change your minds later.
- Break the project's development into (much) smaller chunks. Describe
each chunk thoroughly. What dependencies are there between the pieces? Can
you connect the overall project as a linked schedule or pert chart?
Estimate the estimate the time required to develop each chunk and sum up
the total time along critical paths. Who will work on each chunk?
Most likely, these plans will change as time evolves, but it is a very
useful exercise to plan it out anyway.
- Most groups (I hope) will have some machine-learning or information
extraction component in their system. What will you use as training
data? Does it require a crawler to acquire? How will you label the
data (already labeled? manually by team members? with
crowdsourcing? Distant supervision? What kind?) Will you use the
same source & method for both training and test data?
- What computational resources will you use?
- Describe a series of roughly three milestones you expect to pass as a way of us
each being able to monitor progress at roughly 2 week
intervals.
- Explain how you will evaluate
success. What will you measure, how will you measure it, how will you
display it in the final report? (graphs are usually better than tables if
you can think of a good way to visualize the data)
|
CSE Home 
About Us