Draft Review Form for ICML-2001

Please note that reviewers should use the on-line review form provided by CyberChair, the ICML-2001 conference software. This page is provided only as a sample to give authors information about the review criteria.

Code Number:

Summary. Please summarize what this paper is about in one sentence.

Goals and Tasks. Does the submission state the goals of the research, including the criteria by which readers should evaluate the results? Does it specify clearly the performance and learning tasks under study? How could the authors improve the paper along these lines?

System Description. If the paper describes a system, does it specify the knowledge representation, performance element, and learning algorithm in enough detail to let readers replicate them? How might the authors describe their framework more effectively?

Claims and Evidence. Do the authors make explicit claims or draw clear conclusions, and do they present reasonable evidence to support their position? Do their results reveal the underlying reasons or causes for phenomena? How might they strengthen their claims, evidence, or reasoning?

Context and Limits. Does the submission motivate the research, place it in the context of previous work, and explain its contribution to the literature? Do the authors note their approach's limitations and suggest directions for future work? How could they do better on these dimensions?

Communication. Does the paper communicate well to the reader? Does the manuscript have a rational organization, with understandable sentences, necessary transitions, and correct grammar? Can you suggest ways to improve its readability?

Other Comments. Do you have any other suggestions or comments for the authors that would help them improve their paper or their research?

Recommendation. Should the paper be accepted in its current form, accepted on condition that the authors make the minor presentation improvements suggested above, marginally accepted or rejected entirely? In all cases please provide a sentence or two explaning your recommendation (see below for more complete guidelines for each category).

   Accept [ ]  Conditional Accept [ ]    Marginal [ ]   Reject [ ]  

Accept: great paper, no presentation/methodological flaws.

Conditional accept: great idea, presentation needs improvement (proof has a minor error, another small experiment needs to be run, the writing is not clear). This designation is to be used infrequently.

Marginal: Paper has some good ideas, but would need to be accepted with flaws and only small improvements. Flaws would take longer than three weeks to fix.

Reject: Paper has too many flaws to be accepted this year.

Please send comments about the review form to icml2001@ecn.purdue.edu .