interspeech-reproducibility
Use when hardening the reproducibility of an INTERSPEECH paper — pinning corpus versions and official splits, publishing text-normalization and scoring rules that WER/EER silently depend on, documenting MOS listening-test protocols, reporting seeds and variance within 4 pages, and making toolkit recipes rerunnable.
Profile →npx skills add brycewang-stanford/Awesome-Journal-Skills --skill interspeech-reproducibility --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# INTERSPEECH Reproducibility Speech results decay through the *measurement*, not just the model. Two labs with identical checkpoints can report WERs a point apart because their text normalization differs, and two identical TTS systems can score half a MOS point apart under different listening panels. Interspeech reviewers know this, and the 4-page format means reproducibility is asserted through precise, compressed disclosure — there is no appendix to hide vagueness in. ## The five decay channels | Channel | How results drift | Pin it by | |---|---|---| | Corpus version & splits | "test" ≠ official test; filtered utterances | Name corpus version + official partition; list any filtering rule | | Text normalization | casing, punctuation, numerals change WER | Publish the norm script; name the scorer (e.g., sclite/jiwer + config) | | Trial lists / protocols | EER/minDCF move with the trial set | Cite the exact trial list file and calibration set | | Subjective testing | MOS panels differ in raters, scale, stimuli | Report rater count, platform, instructions, #stimuli, CI | | Training stochasticity | seed, data order, nondeterministic kernels | Seeds logged; ≥3 runs where feasible; me
- The five decay channels
- Corpus discipline
- Objective metrics: publish the ruler
- Subjective metrics: the protocol is the result
- Variance and seeds inside 4 pages
- Recipe rerunnability
- Review-time vs camera-ready disclosure
- The disclosure block, worth three lines of the paper
- Reviewer questions to preempt
- Output format
What does the interspeech-reproducibility skill do?
Use when hardening the reproducibility of an INTERSPEECH paper — pinning corpus versions and official splits, publishing text-normalization and scoring rules that WER/EER silently depend on, documenting MOS listening-test protocols, reporting seeds and variance within 4 pages, and making toolkit recipes rerunnable.
How do I install it?
Run `npx skills add brycewang-stanford/Awesome-Journal-Skills --skill interspeech-reproducibility --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From brycewang-stanford/Awesome-Journal-Skills, a repository with 909 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.