[PRE REVIEW]: Software Design and User Interface of ESPnet-SE++: Speech Enhancement for Robust Speech Processing

[editorialbot]

Submitting author: !--author-handle-->@neillu23<!--end-author-handle-- (Yen-Ju Lu) Repository: https://github.com/espnet/espnet Branch with paper.md (empty if default branch): Version: v.202207 Editor: !--editor-->@faroit<!--end-editor-- Reviewers: @joimort, @justusschock Managing EiC: Kristen Thyng

Status

Status badge code:

HTML: <a href="https://joss.theoj.org/papers/f518d0b78a29ab6521936834e5d90ef1"><img src="https://joss.theoj.org/papers/f518d0b78a29ab6521936834e5d90ef1/status.svg"></a>
Markdown: [![status](https://joss.theoj.org/papers/f518d0b78a29ab6521936834e5d90ef1/status.svg)](https://joss.theoj.org/papers/f518d0b78a29ab6521936834e5d90ef1)

Author instructions

Thanks for submitting your paper to JOSS @neillu23. Currently, there isn't a JOSS editor assigned to your paper.

@neillu23 if you have any suggestions for potential reviewers then please mention them here in this thread (without tagging them with an @). In addition, this list of people have already agreed to review for JOSS and may be suitable for this submission (please start at the bottom of the list).

Editor instructions

The JOSS submission bot @editorialbot is here to help you find and assign reviewers and start the main review. To find out what @editorialbot can do for you type:

@editorialbot commands

[neillu23]

Hi @faroit! Unfortunately, I don't know of that many people who don't have conflicts with the institutions on the paper. @popcornell, @simpleoier, do you know of anyone who can help us with the review?

[oliviaguest]

@neillu23 maybe check https://reviewers.joss.theoj.org/ too? 😊

[faroit]

@klukosiute @shruti-singh @vimalthilak @levimcclenny @Het-Shah @NMontanaBrown @pmeier @justusschock @yunusrf @JesseLivezey @ethanwharris @paragkulkarni11 @joimort

The software under review is espnet is one if the most used frameworks in the field of audio machine learning based audio/speech processing – would any of you be willing to review this submission for JOSS? Domain knowledge is a plus but not required for this review.

The JOSS review process takes place on GitHub and focuses on the software and a short paper. We carry out our checklist-driven reviews here in GitHub issues and follow these guidelines: https://joss.readthedocs.io/en/latest/reviewer_guidelines.html

This issue is a "pre-review" issue in which reviewers are assigned. Once sufficient reviewers are recruited we will open a dedicated review issue where the review will take place.

[pmeier]

I can review, but can't offer any domain knowledge.

[klukosiute]

I can also review but can’t claim to offer very close domain knowledge.

[justusschock]

@faroit my domain knowledge is also very limited. If that's fine for the three of us, I can review as well!

[NMontanaBrown]

@faroit Same as other reviewers, I am happy to review but can't claim domain knowledge.

[joimort]

@faroit Same here, and I do have a background in audio processing and speech enhancement. So, I would be happy to review and provide feedback!

[faroit]

@pmeier @NMontanaBrown @klukosiute thanks for your response and the willingness to review!

Given that we have a reviewer with audio background (thanks, @joimort for accepting) now, I would like to ask @justusschock to review this and bring in his experience with torchmetrics which might help with espnet as there is some overlap when it comes to speech separation metrics.

[faroit]

@editorialbot add @joimort to reviewers

[editorialbot]

@joimort added to the reviewers list!

[faroit]

@editorialbot add @justusschock to reviewers

[editorialbot]

@justusschock added to the reviewers list!

[faroit]

@neillu23 before we can (finally) start the review process, would it make sense to update the version of espnet under review to 202301?

Also reading the paper again, i think it would be good to explicitly mention the specific PRs in the footnotes that were merged for espnet se++ so that its easier to see the code contributions.

[faroit]

@neillu23 please update us on the versioning question above

[neillu23]

Hi @faroit, we can upgrade to version 202301, which includes all the features of ESPnet-SE++. However, it's worth noting that most of the updates in this new version aren't related to SE. Regarding code contributions, the ESPNet-SE++ paper provides instructions and software structure for both SSE and Joint-Task work. It might be challenging to pinpoint specific SE PRs since all SE closed PRs are likely included. Instead, it would be easier to recognize Joint-Task PRs like the "enh_s2t joint model #4226" and the "LT-Spatialized and SLURP-Spatialized combined enhancement recipe #4268." We can update this information in the footnotes.

[faroit]

@neillu23 well noted. I think a few comments with respect to the related PRs in the footnotes might helpful. In any case, lets finally start the review then. Sorry for the long wait

[faroit]

@editorialbot start review

[editorialbot]

OK, I've started the review over in https://github.com/openjournals/joss-reviews/issues/5403.

[neillu23]

@faroit thank you very much for the help! I will add comments on the footnotes