Referee Guidelines


ACM TSLP recognizes that reviewing is a service to the profession; this publication endeavors to treat reviewers with courtesy and respect. ACM guarantees certain rights to reviewers. TSLP provides reviewers additional guarantees.
Papers for the ACM Transactions on Speech and Language Processing (TSLP) must be of high quality and fall within the scope of the journal. There are three main ingredients to an acceptable paper.

  • The technical quality is high.
  • Interest and novelty is high.
  • The presentation is effective.

Few papers excel in all of these, but a substandard level in any is sufficient ground for rejection. Many papers require substantial revisions before acceptance, and reviewers should not hesitate to recommend that a paper be rejected pending changes that are required for completeness, correctness, or to substantially improve clarity.
TSLP places high emphasis on wide accessibility of papers. TSLP strongly encourages authors to include examples where appropriate and to make greater efforts to target their presentation to a broader audience than specialists doing current research in the topical areas of the papers. Please determine if the paper is readable. If it is not, suggest how it may be improved (e.g., by requesting illustrative examples, expanded discussions on key points that are not clear, etc). Please determine that the submission situates the current work with respect to published work by providing a reasonable summary of the latter.
More specific criteria apply to papers presenting experimental results. In such papers, authors should strive to report experiments with replicability as a goal. Such papers should report results on standard test sets using standard metrics. Authors should cite the best known results on these test sets. Authors should provide statistical significance tests on their results. There is little value in a paper that describes an experiment using authors' private data, private test sets, and authors' own metric. In case there are no standard tests or metrics in the paper's area, the authors should have a mechanism to provide the test material and the evaluation tool to the community. Please determine that the paper you review satisfy these desiderata.
Please verify that papers advocating a new algorithm report the best baseline that can be obtained using known alternate methods. Please verify that papers advocating automatic methods make clear all the manual steps (if any) involved.
The following is a list of other considerations to be taken into account when reviewing a submission.
Please treat the material in a submitted paper as confidential; please make every conscious effort not to use or build upon the material in the paper until it has seen publication.
Consistent with the ACM Policy on Reviewer Anonymity, reviewers must maintain the confidentiality of reviewer identities, as well as the reviews themselves, that are communicated to them at any time.
TSLP will publish outstanding papers which are "major value-added extensions" of papers previously published in conferences; that is, TSLP will not automatically reject papers that are substantial extensions of previously published conference papers. These papers will go through the normal review process. The submitted manuscript should thoroughly consolidate the material, should extend it to be broader, and should more carefully cover related research. It should have at least 30% new material. The new material should be content material, not just the addition of proofs or a few more performance figures. This affords an opportunity to describe the novel approach in more depth, to consider the alternatives more comprehensively, and to delve into some of the issues listed in the other paper as future work.
TSLP would like to discourage excessively long papers (longer than 50 double-spaced pages including figures, references, etc.) and unnecessary digressions even in shorter papers. This is to help the authors to focus on the most important aspects of their submission, to make it easier for the reviewers and readers, and to allow more papers to be published in any given issue. Please determine if the paper you review can be shortened without materially compromising the completeness and worthiness of the paper.
In a similar vein, TSLP encourages shorter submissions, including even very short (say, five page) submissions. The primary criterion for acceptance is improving on the state-of-the-art in some significant way.
TSLP also publishes focused surveys. These should be deep and will sometimes be quite narrow, but would make a contribution to our understanding of an important area or subarea of speech and language processing, broadly defined. TSLP surveys should be educational to speech and language audiences by presenting a relatively well-established body of database research. Surveys can summarize prior literature on a theoretical or systems research topic, or can explain approaches implemented in commercial systems. A survey of the former type summarizes a literature on a particular subject, presenting a new way of understanding how the papers in this literature fit together. A survey of the latter type summarizes the best industrial art, and can be acceptable even if it represents no new contribution over what has been used in industry for years, if the paper's content is not to be found in the published literature.

About TSLP

  • Charter and Scope
  • Editorial Board
  • Special Issues

For Authors

  • Manuscript Preparation
  • Submission Instructions
  • After Acceptance
  • Conflict of Interests
  • Appeals Procedure
  • ACM Policy

For Readers

  • Current Issue
  • Archive

For Referees

  • Rights and Responsabilities
  • Guidelines

Related ACM Links

  • ACM Digital Library
  • ACM Web Site