CSCI 544 Course Project

Spring 2015
Kenji Sagae

Project Proposal

Due March 25

• What is the problem or task addressed in the project?
• Why is it interesting, and why is it challenging?
• Who would benefit from this work, and how?
• What data and knowledge sources will be used?
• How will data be collected and annotated (if applicable)?
• What specific techniques do you expect to use? (Answer if possible. This is not a firm commitment, especially for projects on topics not yet covered in class, e.g. dialogue)
• How will the approach be evaluated?
• What does each member of the team plan to do? (Subject to change!)

Report

Due May 1

In general, the final report should cover the general motivation and background of the work, related work, technical details, evaluation (or a detailed evaluation plan), conclusions and possible directions for future work. Although details will differ from project to project, grading will be based on the following general guidelines:

• Introduction, motivation and general description of the problem (10%)
  • What is it the problem you are solving?
  • Why is it interesting?
  • Who would use it when solved? And how would it be used?
  • Why is it challenging?
  • What are the shortcomings and limitations of the existing work?
• Related work (10%)
  • How does your work relate to those done by others in the NLP field?
  • Provide a citation to the papers you have read, explain briefly what each paper is about, what are the pros and cons of the approach, how does it compare and contrast to your approach? Did your method improve existing algorithms on the same dataset?
• Data (including collection and annotation, if applicable) (20%)
  • Describe the data you have used (size, origin, and other relevant details)
• Technical approach (25%)
  • Provide a detailed description of your approach (including algorithms, features, etc.) There is no need to describe well known techniques, such as Naive Bayes classification or the Perceptron algorithm, in detail.
• Evaluation and analysis (25%)
  • Describe your evaluation, or a detailed plan for evaluation, including specific metrics, datasets, users, etc.
  • How does your approach compare to a baseline system?
  • How does your approach compare to comparable work?
  • What does your system get right? What does it get wrong, and why?
• Style and writing (10%)
  • Is the writing clear? Is the report organized well? Are there obvious errors?
  • Could an instructor, a classmate, or an NLP researcher understand exactly what you did, and possibly replicate your findings, based on your report?

Reports should be formatted following the guidelines from the Annual Meeting of the Association for Computational Linguistics. Students should use the MS Word template or the LaTeX style files provided in the ACL 2014 call for papers.

Reports should be four to eight pages long, including figures. Each team member must contribute with writing, and individual contributions should be clearly marked. One extra page is allowed for references, if necessary.

Brief report on other projects

In addition to the project report (only one per team), each student must also turn in an individual report describing 10 projects by other students. For each of the 10 projects, include the team number and project title, and write a few sentences describing the project. Information you may wish to cover include: the main idea, a brief statement of the technical approach, and strengths/weaknesses of the project. This must come from information you gather by visiting other teams during the poster sessions. The entire report must be between two and three pages. This is an individual report to be turned in in addition to the team project report.