Linguistics 473
Computational Linguistics Fundamentals
Summer 2017

Writing Assignment


Assignment 5: Critical Paper Review

Due: Tuesday September 5, 2017, 11:45 p.m.

The main purpose of the assignment is to give you practice in reading and understanding research papers and thinking critically about them. For this assignment, you will write a critical review of a published, peer-reviewed paper from the computational linguistics literature. You may select one of the papers below, or pick a relevant paper on your own (instructor approval).

Because academic work must adhere to strict guidelines for journal publication, proper formatting is a skill you will need for research, employment, and for many classes in the UW computational linguistics cirriculum. Accordingly, for this assignment, proper formatting is required and is a graded criteria for this assignment.

Format your paper according to the ACL-2017 guidelines. Templates for LaTeX and Word are located at http://acl2017.org/calls/papers/ (see the "Paper Submission and Templates" section for LaTeX and Word templates). Your review should have an abstract and be at least two pages according to the ACL template. Cite the paper you are reviewing, plus any other papers you mention, in a References section.

The paper review should contain:
  • A summary of the paper, detailing:
    • the problem addressed,
    • the approach,
    • the main results.
    Note: this should be a summary and not a direct copy of the text of the paper. You can copy figures and tables from the paper.
  • A critical analysis, addressing the following points:
    • Did the authors explain the problem and review the prior work or the state-of-the-art techniques for the problem?
    • Do the experimental results convincingly demonstrate the benefit of the approach? (e.g. have the authors used a reasonably strong baseline and fair experimental conditions).
    • What are the limitations of the approach? (e.g. will it work on other languages and in other settings (more or less data, availability of annotated resources, constraints on running time or memory consumption))
    • Are there other possible applications of the approach (e.g. can we address other phenomena using a similar approach)?
    • What are possible directions for future work building on this approach or addressing the same problem in a better way?
    Don’t worry too much about the prose quality. The main goal is to see whether you understand the paper and can think critically about it. Submit your critical paper review to your CollectIt dropbox in PDF format.

    Grading

    Content - 80%
    ACL Formatting - 20%

    Acceptable papers

    You can choose any relevant, peer-reviewed or published journal paper from any subfield of computational linguistics, including but not limited to the following:
    • Named entity recognition
    • Word sense disambiguation
    • Anaphora resolution
    • POS tagging
    • Clustering
    • Classification
    • Parsing
    • Question answering
    • Evaluation metrics
    • Computational semantics
    • Computational linguistic typology
    • Lexicon induction
    • Machine translation
    • Information retrieval
    • Document summarization
    • Information extraction
    • Clustering
    • Speech Recognition
    • Speech Synthesis
    Any paper delivered at a major conference (ACL, COLING, EMNLP, NAACL, etc.) can be used. Important papers from regional conferences can also be used. If you have questions about the eligibility of the paper, please consult the instructor for approval.

    If a paper that you find via web searching has a fee-based download, the paper might be available online through the UW libraries. For this, you must be logged in to UW Libraries with your UW NetID.