Last Modified: January 1, 2017
CS 295: Statistical NLP: Winter 2017
Writing Paper Summaries
Sameer Singh
As part of the learning objectives of the course, I expect each student to be able to read, understand, and
critically analyze recent research papers in the field of natural language processing. To aid this, each student will
have to submit three research paper summaries during the course (of course, you are encouraged to read other
papers as well). This write up describes what I expect from each submission.
Each paper summary will be a
PDF write-up containing two sections: Content Summary and Analysis.
Refer to Canvas to see which papers are assigned to you, and when they are due. The papers are available on the
course website at
. I do not plan
to change the paper assignments, however, if you feel really strongly that you are completely unprepared to review
the paper (for example if you do not have a computer science background or are an undergraduate student), send
me an email arguing your case and propose alternate papers from the list that you would rather review.
1 Content Summary
In this section, you should present a summary of the main content of the paper, in your own words. You should try
to be fairly objective here, you are just presenting a shorter version of the authors’ arguments. Your summary
should address the following issues (approximately 1-2 sentences each):
1. Motivation: describe the general problem that the authors are focusing on, and why it is important,
Contributions: what have the authors actually proposed as the solution, and why they think it is a good idea,
3. Related Work: what have others done, why those don’t work, how does this work differ, and
4. Support: what evidence have the authors provided to demonstrate that their proposed ideas work.
The above is not a fixed structure, feel free to reorder them or not follow these subsections at all, but these four
aspects of the paper should be evident in your summary.
2 Analysis
In this section, you should provide your own views on the paper, both positive and negative, along with suggestions
to improve the work. There are a number of issues/questions you should keep in mind when writing the analysis
(this is neither a comprehensive nor a necessary list of things to cover):
Writing: what did you think of the quality of writing? amount of background covered? organization of the
sections? clarity in describing the technical aspects? how would you improve it?
Motivation: do you think this is an important problem? should more people be looking at this? why?
Contributions: were the contributions interesting? novel? incremental?
Evaluation: was there sufficient empirical evidence to support the claims? are there other experiments they
could have done? did the experimental setup have flaws that can be addressed?
Future Work: what do you think should be a good follow-up work? which new threads of research open up
because of this? have the authors provided resources to support future research (code/data)?
3 Statement of Collaboration
Optionally, include a Statement of Collaboration. Every student is required to adhere to UCI’s academic honesty
guidelines, available on the course homepage. I encourage students to reach out to others reviewing the same paper
using Piazza, and encourage free discussions of the papers (online, as much as possible). However, everything in
the submission should be written by you, and as a general guiding principle, you are not allowed to take anything
physical away from the discussions (so no photographs, written notes, referring to Piazza constantly, etc.). When
in doubt, include the statement describing the extent of your discussions and collaborations.
Paper Summaries UC Irvine 1/ 1