All deadlines are on Canvas.

In-Class Assignments (25 pts each)

We’ll have two class sessions dedicated to in-class activities or “homework”. These assignments will be paper and pencil.

During these class sessions, don’t:

Use open electronics (cell phones, laptops, Meta glasses, etc), unless needed for accommodation purposes.
“Crowdsource” answers across the entire class and broadcast them in a lecture-like manner.

You may instead:

Collaborate with your classmates and help each other build understanding.
Use notes you hand-wrote.

You should:

Show step-by-step work and/or justification for your thinking on your own assignment to receive credit.
Write the names of people you worked with on the first page.

In-class assignments are due at the end of class. If you are unable to make the date for an in-class assignment, you may complete an alternative assignment during a separate date and time in Lucy’s office.

You cannot submit a regrade request multiple times for the same problem.

Research Project (70 pts)

The aim of this project is to produce a paper worthy of submission to a workshop or conference.

Team size. These are done in groups of 2-4.

Deadlines. There are no late days – please plan ahead and coordinate with your group members if one of you has a conflict. Each day late incurs a 10% point penalty. Late days penalties are applied to assignment grading at the end of the semester. We follow Canvas’s implementation of time for determining whether assignments are submitted late. It is possible that a deadline of 11:59 pm on Canvas is 11:59:00 pm and not 11:59:59 pm, and so you should avoid submitting too last-minute to avoid worrying at this level of granularity.

Grading. Grading for the research project is shared across group members. We may adjust individual members’ grades in extreme cases of unequal contribution.

Project Overlap. You can work on a project that has external collaborators who are not students in the course, but you must note this in your final report and delineate what was your work and what was theirs. You can share a single project between our class and another class, but you must declare that you are sharing the project in your project proposal, and the expectation is that your contribution and effort will be bigger.

Compute. On the first day of class, we will discuss compute opportunities, e.g. instructional GPUs and Google Cloud Credits.

Policies. You are allowed to use existing codebases from prior work as long as you document it. You’ll be graded based on what you contribute. Using AI to generate paragraphs of written content in your proposal or final report is considered academic misconduct. Writing enables active thinking, and we would like you to actively think through your project as a human. Other forms of AI assistance (e.g. programming, generating figures, making writing edits, and surfacing related work) are allowed. Overleaf revision history for the proposal and final report may be used for resolving potential contribution disputes within project groups.

Brainstorming (5 pts)

We will have one class session dedicated to brainstorming presentations.

Propose a project idea. This live pitch should include why the question/s being asked are important and why your project can be completed feasibly within this semester.

Research question (1 pt). This question should be scoped so that it is addressable within the weeks we have remaining in this class.
Core contribution (2 pt). You should juxtapose your contributions against relevant prior papers, e.g. what do you add that past papers do not? Are you working on a new or spin-off problem that has not received the attention it deserves? Cite representative prior work.
Method (1 pt). What models and tasks? Why these methods?
Data (1 pt). You can find datasets by reading papers that contribute metadata-rich resources or benchmarks. You may also create a dataset from scratch, though it’s possible that doing so may require most of the semester’s time and become your main contribution (e.g. a resources and benchmarks paper).

You will pitch your slide in 1-5 minutes to the class, exact time TBD depending on the number of project groups we have. You will submit one Google Slide, and we will compile all slides into one mega slide deck (taking the version of the document link closest to the deadline). If you have presenter notes, have those with you, as we’ll be running the mega presentation on a laptop on our end.

Peer Feedback I (5 pts)

A folder of slide decks of groups’ brainstorming efforts, locked to editing but open to comments, will be shared with the class. You are expected to add comments on each other’s projects to give at least three other project teams feedback, ideas, or questions. This will be graded as all or nothing. Grading will not only consider just the number or presence of comments, but the quality of comments. Please spread out your comments so that each team gets some feedback from someone.

Proposal (10 pts)

Expand on your project idea from the brainstorming phase (700-1000 words). It is okay if your proposal deviates from your brainstorming presentation, as that is a natural side effect of becoming wiser.

Research question and contributions, scope appropriately. (2 pt)
Literature review. (3 pt)
Your data and approach, including potential experiments and pipelines needed to carry out the project. (5 pt)

Write your proposal directly in Overleaf as a human using the ACL conference paper template, and include the link to the Overleaf as a footnote in the PDF your submit onto Canvas (with permissions set to “anyone with the link can view”).

Midterm Presentation (10 pts)

We will have two class sessions dedicated to midterm presentations.

This presentation covers progress made so far, data collected, challenges faced, preliminary results, and the timeline needed to complete the project. Your presentation should cover related work, your dataset, your methodological approach, preliminary results (e.g. figures and tables), and next steps. You may also discuss ways in which your new goals have deviated from your original ones.

You will be graded based on:

presentation delivery and clarity. (3 pt)
whether your progress is substantive. (5 pt)
how well you handle questions posed by the audience. (2 pt)

It is okay to not have all of the answers, but you should be able to explain details about your process and project. If you use AI to make key decisions in your project, you should be able to explain why the decisions being made are good decisions.

The length of this presentation is TBD based on the number of project groups. You will submit your slides on Google Slides, and we will compile them into a single mega-slide deck to ease presentation timing.

Peer Feedback II (5 pts)

The format and grading of this matches our peer feedback stage for brainstorming slides.

Final Report (25 pts)

Project teams submit their papers as PDFs with Overleaf links (8 pages).

Abstract (2 pts): Motivate the problem, highlight main goals, include any preliminary main findings.
Code repository (1 pt): A link to a public Github and/or Huggingface respositories.
Research question/s (2 pts): Try to limit this to less than three.
Related work (2 pts): Recap related work and indicate what is novel in your work compared to prior work.
Data (4 pts): Recap your dataset and basic statistics about it (e.g. total word count, number of sentences, metadata information).
Approach (5 pts): If your computational project advances a task methodologically, include baselines, parameter decisions, and evaluation. If you are running an experiment or applying a pre-existing method, include details on parameters and experimental design. Motivate the validity of your methodology.
Results (8 pts): A written section that references key figures and tables, which include captions that are descriptive enough to stand alone. Grading for the results section depends on whether it shows substantial progress, careful consideration of others’ midterm feedback, and fulfillment of goals since the midterm report. Failure to meet goals is okay if there is thoughtful discussion around what was challenging and attempts you made towards success, and negative results are also acceptable especially if accompanied by explanation and/or additional analysis. Your results should be sound; you should support your claims with clear arguments and evidence.
Conclusion and Future Work (1 pts): Provide a summary of your work and its implications, as well as follow-up potential work.
Contributions (0 pts, but if missing, -1 pt): Describe the contributions of each collaborator (including AI use) across your project team.
Formatting (0 pts, but -1 pt possible): One penalty point will be applied if the PDF you submit does not pass aclpubcheck. The penalty in actual research settings is harsher; papers get desk rejected when formatting rules are not followed.
Clarity (0 pts, but -5 pt possible): Up to five penalty points may also be applied if your writing is unclear or if the paper’s content is poorly organized.

Like with the project proposal, please include a link to your pdf Overleaf as a footnote in the PDF you upload onto Canvas. Directly write your report in LaTeX in Overleaf.

⚠️ The due date for the final paper is very close to the due date for the poster, to anticipate potential last minute changes to your results during poster preparation. We highly recommend starting and finishing this paper early!

Poster Presentation (10 pts)

Our final class session will involve a poster session, open to others in the CDIS community and beyond.

Grading:

Live question responses (5 pt). You should provide thoughtful answers to questions the instructor/TA asks about your work and the decisions you made. Your answers should demonstrate engagement and understanding with the details of your work.
Poster PDF (5 pt). Your poster should be clear and well-organized.

Academic Misconduct

To respond to cheating and plagiarism, we will follow policies and procedures determined by UW-Madison.