# CS6741 Project Requirements and Guidelines

There are two project options: research and survey. If you are doing research in NLP, the first option is probably the best fit. If you are relatively new to the field, a literature survey is a great way to dive in to your favorite topic. If it helps, please email the instructor to discuss your preference.

Regardless of your choice, each project will include: (a) a proposal abstract, (b) a proposal presentation, © final presentation, and (d) a final report.

## Option 1: Research Project

You will do a classical research project to address an NLP problem. We take a broad view of the field, but you should confirm your topic with the instructor. Your research will study an approach to address the problem and will include rigorous empirical and qualitative evaluation. Keep in mind the relatively short time available (less than one semester) and available data when considering the feasibility of your project. Using your research, if relevant to the topic of the course, for your project is recommended. If your research project’s scope is beyond a single semester, please identify intermediate goals to complete for this project.

## Option 2: Literature Survey

You will survey at least 20 papers and identify research trends and progress. The survey must include a diverse set of papers starting from when the problem was studied in NLP. You will also need to identify and discuss related papers in linguistics. You will identify a coherent trajectory and describe how it aligns with related trends in the field. Your goal is to acquire a deep understanding of the problem and develop your paper reading skills. Your discussion must encompass resources, methods, evaluation, and problem setup.

## Presentations and Reports

The dates for the presentations are listed on the course website. The project due date will be shortly after the end of the semester. The exact date will be announced.

### Proposal Abstract

Abstracts due: October 2, 11:59pm on CMS

Please submit your choice, title, abstract, and other required material (see below) by the deadlines listed or announced later in the semester. The abstract is limited to 300 words (give and take some prepositions). Please use the abstract to state your problem, general approach, data sources, and proposed experiments (just like a paper abstract).

If you are doing a research project, please also submit a separate paragraph (up to 300 words) describing your work on the project so far. If you are doing a literature survey, please also submit an initial reference list with at least 10 papers that you plan to include in your survey.

### Proposal Presentation

Presentations scheduled to October 23 and 25. Slides are due October 22, 5pm on CMS

Please describe your project and defend it. Please submit your slides in a single PDF file. Each presentation will be 10 minutes, followed by questions.

Some guidelines:

• Argue why your problem is important
• Argue why your problem is challenging
• Argue why your approach is reasonable
• Describe what you did so far
• Show us your data: use examples for all parts of your presentation
• If your task is multi-disciplinary (e.g., language+vision), describe what aspects of language will play a role Description and motivation of your technique and method
• Discuss related work that inspired you and/or justified your approach
• Discuss feasibility given available resources (relying on prior work is valid)
• Identify at least one corpus you will use (more is better!)
• It is recommended that you use examples and figures to illustrate the problem and your approach (I can’t emphasize this enough)
• If you plan to collect data, describe the process and do a quick feasibility study
• If you are doing a literature review, show the overall structure of how research developed in the field and give us a glimpse at 2-3 key approaches

### Final Presentation

Presentations scheduled to November 27. Slides are due November 26, 8pm on CMS

Each presentation should be 10 minutes. I will stop you are after 11 minutes. The presentation will be followed by questions.

It is recommended that you use examples and figures to illustrate the problem and your approach. Please present your approach, experiments and result. Do not assume the class remembers the details of your problems or approach, although you can dedicate less time to the related work. The proposal presentation guidelines are applicable to the final presentation as well.

### Final Report

Due: December 8, 11:59pm on CMS

The final report is limited to 5 pages in ACL format, plus unlimited pages for references. The report should be written as a research paper or a literature survey. If you are not doing a survey, please also submit your source code in a zip file.

This is a short report, so organize your content well. It is recommended to fashion your report after good short papers submitted to ACL/NAACL/EMNLP. There are plenty of excellent examples!

For the research track, it is recommended to include at least all the following in your report:

• Introduction, including problem definition and motivation
• Examples, examples, examples!
• Related work
• Technical background, if required (No, you don’t need to introduce SVM or Back-prop)
• Formal well-motivated description of your approach
• Experimental setup
• Results
• Error analysis

Ng 2010 is a great example of a literature survey.