Stored boundaries on the spectrogram

25 windows from the eleven hand-segmented corpora, one to two files per corpus, the 2.5 s with the most vowels in each file. Every line is a boundary exactly as stored in the dataset; the formant dots are re-measured with the dataset's Praat settings, so they show what the frames tables hold.

vowel (gold lines, tinted tier) consonant / pause boundary F1 F2 F3word tier above the spectrogram

English

Buckeye · L2-ARCTIC · TIMIT · 6 windows

English s0301a spectrogram with boundaries — Buckeyes0301as03aligner output, hand-corrected1461 stored phones in file

English s0403a spectrogram with boundaries — Buckeyes0403as04aligner output, hand-corrected2688 stored phones in file

English EBVS_arctic_a0032 spectrogram with boundaries — L2-ARCTICEBVS_arctic_a0032EBVSaligner output, hand-corrected31 stored phones in file

English LXC_arctic_a0117 spectrogram with boundaries — L2-ARCTICLXC_arctic_a0117LXCaligner output, hand-corrected25 stored phones in file

English MDRB0_SX184 spectrogram with boundaries — TIMITMDRB0_SX184MDRB0segmented by hand9 stored phones in file

English MGWT0_SX9 spectrogram with boundaries — TIMITMGWT0_SX9MGWT0segmented by hand20 stored phones in file

Arabic

Arabic Speech Corpus · 1 window

Arabic main-ARA_NORM_1179 spectrogram with boundaries — Arabic Speech Corpusmain-ARA_NORM_1179halabi_speakeraligner output, hand-corrected30 stored phones in file

Basque

VoxAngeles · 1 window

Basque eus-001-014 spectrogram with boundaries — VoxAngeleseus-001-014eus-001aligner output, audited9 stored phones in file

Central Khmer

VoxAngeles · 1 window

Central Khmer khm-000-045 spectrogram with boundaries — VoxAngeleskhm-000-045khm-000aligner output, audited5 stored phones in file

Dutch

IFA · 2 windows

Dutch F20N1FT7F spectrogram with boundaries — IFAF20N1FT7FF20Nsegmented by hand37 stored phones in file

Dutch F28G1FPA1VCVE spectrogram with boundaries — IFAF28G1FPA1VCVEF28Gsegmented by hand21 stored phones in file

Finnish

VoxAngeles · 1 window

Finnish fin-002-124 spectrogram with boundaries — VoxAngelesfin-002-124fin-002aligner output, audited12 stored phones in file

German

PaVoQue · 1 window

German neutral-a0880 spectrogram with boundaries — PaVoQueneutral-a0880roettigaligner output, hand-corrected57 stored phones in file

Gujarati

VoxAngeles · 1 window

Gujarati guj-000-013 spectrogram with boundaries — VoxAngelesguj-000-013guj-000aligner output, audited7 stored phones in file

Hausa

VoxAngeles · 1 window

Hausa hau-010-009 spectrogram with boundaries — VoxAngeleshau-010-009hau-010aligner output, audited5 stored phones in file

Hungarian

VoxAngeles · 1 window

Hungarian hun-001-042 spectrogram with boundaries — VoxAngeleshun-001-042hun-001aligner output, audited6 stored phones in file

Korean

Seoul Corpus · 2 windows

Korean s32m43m3 spectrogram with boundaries — Seoul Corpuss32m43m3s32aligner output, hand-corrected5792 stored phones in file

Korean s35m43f2 spectrogram with boundaries — Seoul Corpuss35m43f2s35aligner output, hand-corrected4386 stored phones in file

Lakota

VoxAngeles · 1 window

Lakota lkt-000-039 spectrogram with boundaries — VoxAngeleslkt-000-039lkt-000aligner output, audited9 stored phones in file

Modern Greek

VoxAngeles · 1 window

Modern Greek ell-005-045 spectrogram with boundaries — VoxAngelesell-005-045ell-005aligner output, audited9 stored phones in file

Persian

Persian Speech Corpus · 1 window

Persian 360-A spectrogram with boundaries — Persian Speech Corpus360-Ahalabi_speakeraligner output, hand-corrected120 stored phones in file

Swedish

Waxholm · 2 windows

Swedish fp2007.8.06 spectrogram with boundaries — Waxholmfp2007.8.06fp2007aligner output, hand-corrected17 stored phones in file

Swedish fp2041.14.04 spectrogram with boundaries — Waxholmfp2041.14.04fp2041aligner output, hand-corrected44 stored phones in file

Thai

CCOST · 2 windows

Thai ccost_S34_INT_4 spectrogram with boundaries — CCOSTccost_S34_INT_4S34aligner output, hand-corrected834 stored phones in file

Thai ccost_S7_ST_123 spectrogram with boundaries — CCOSTccost_S7_ST_123S7aligner output, hand-corrected611 stored phones in file

Gaps in the phone tier are phones that belong to no V(C)V sequence (vowel hiatus, utterance edges, sequences broken by a pause): the dataset only stores segments that take part in a sequence, so nothing is drawn there. Boundaries are the corpora's own; nothing has been re-aligned.