Common Voice Scripted Speech 26.0 - Malaysian English

Description

This dataset is a derived, accent-filtered subset of the Mozilla Common Voice Scripted Speech 26.0 English release (`cv-corpus-26.0-2026-06-12`, locale `en`). It keeps only validated rows (present in the source `validated.tsv`) whose self-declared `accents` field is *exactly* `"Malaysian English"`. No gender or age filter is applied. The result is split speaker-disjoint into train/dev/test (no `client_id` appears in more than one split).

Specifics

Licensing

Creative Commons Zero v1.0 Universal (CC0-1.0)

https://spdx.org/licenses/CC0-1.0.html

Considerations

Restrictions/Special Constraints

N/A

Forbidden Usage

- You agree not to attempt to determine the identity of speakers. - You agree not to rehost this dataset.

Processes

Splits and speakers: 2,476 validated clips total — train 2,343 clips / 100 speakers, dev 62 clips / 5 speakers, test 71 clips / 9 speakers. Split is speaker-disjoint (no client_id appears in more than one split), proportioned ~94.63/2.50/2.87 by clip count via a greedy per-speaker balancing algorithm targeting 95/2.5/2.5 (with only 114 speakers, achieved ratios deviate more noticeably from target than in larger subsets — speakers have uneven clip counts under a speaker-disjoint constraint).

Clip duration: 3.56 hours total (12,827 seconds) across 2,476 clips. Per-split: train 3.39h, dev 0.08h, test 0.10h. Per-clip: mean 5.18s, median 5.04s, min 1.22s, max 16.90s, std 1.77s.

Gender distribution: male_masculine 1,624; female_feminine 558; blank/unspecified 294.

Age distribution (shipped subset, self-declared): twenties 1,315; thirties 370; blank/unspecified 287; fourties 180; teens 170; fifties 132; sixties 22.

Closed-form accent matching (QA): the accents field is pipe-delimited multi-select in the source corpus, but this subset was deliberately filtered to the closed-form (exact-match) rule — 100% of shipped clips have accents == exactly "Malaysian English", no co-occurring tags.

Open-form vs. closed-form comparison (at execution time): open-form (accents contains "Malaysian English") = 2,580 clips / 122 speakers / 3.74h; closed-form (accents == exactly "Malaysian English") = 2,476 clips / 114 speakers / 3.56h. Closed-form drops 104 clips (4.03%).

Full-corpus accent x gender and age x gender cross-tabs are in stats.md.

Derivation methodology: derived from the Mozilla Common Voice Scripted Speech 26.0 English release (cv-corpus-26.0-2026-06-12, locale en). Filter: validated rows only (present in the source validated.tsv), whose self-declared accents field is exactly "Malaysian English" (closed-form/exact match). No gender or age filter applied. Speaker-disjoint train/dev/test split via a greedy per-speaker assignment balancing toward 95/2.5/2.5 targets. Only the referenced MP3 clips were extracted from the ~2.58M-clip, ~89GB source archive (not the full archive) via a single sequential streaming pass over the source tar.gz.