audio audioduration (s) 2.77 2.03k | text stringlengths 61 42.9k | speaker_id stringlengths 4 32 |
|---|---|---|
i've also had some meals that(2) make me want(2) to dry heave {BREATH} so {BREATH} it's about choosing the parts of the bible about compassion {BREATH} about tolerance about loving your(2) neighbor {BREATH} as opposed to the parts {SMACK} <sil> {UH} about <sil> homosexuality is a sin {UH} or intolerance or(2) violence... | AJJacobs | |
we have indeed {NOISE} taken <sil> the best part of {NOISE} the meat <sil> so let's(2) look today <sil> at a set of photographs <sil> of a people who lost {UH} so that we could {NOISE} gain {COUGH} <sil> and(2) know(2) that when(4) you see these people's(3) faces <sil> that(2) these are(2) not just(2) images(2) of the... | AaronHuey | |
{NOISE} and then as(2) the(2) ammonia re(2) evaporates and combines with the water <sil> back on the erstwhile hot side {BREATH} it creates a powerful cooling effect {BREATH} so it was a great idea {COUGH} that(2) <sil> didn't work at all {SMACK} it blew up {UM} because(2) using ammonia you get {SMACK} hugely high pre... | AdamGrosser | |
<sil> we came up with(2) {UH} a list {COUGH} of <sil> they wanted <sil> band integration(2) {SMACK} that is the machine {COUGH} acting upon the band members {BREATH} specifically not the(2) other way around(2) {BREATH} they wanted the machine action {COUGH} to(3) follow the song {NOISE} feeling so as the song {NOISE} ... | AdamSadowsky | |
and(2) i thought to myself <sil> wouldn't it be great {COUGH} if i had my own <sil> dodo skeleton <sil> and(2) so <sil> i {NOISE} want to point out here at this point that(2) {BREATH} my life {UH} obsessed {BREATH} by {NOISE} objects and the stories that(2) they tell {BREATH} and(2) <sil> this was the very latest one(... | AdamSavage | |
but diagnosing a brain disorder <sil> without actually(2) {NOISE} looking {NOISE} at <sil> the brain is analogous to {SMACK} treating {UH} a patient with a heart {UH} problem based on their physical symptoms without(2) {BREATH} even <sil> doing an {NOISE} ecg {NOISE} or(2) a chest x ray {SMACK} to(2) {NOISE} look at t... | AditiShankardass | |
for(2) better {NOISE} or(2) worse we kids aren't hampered as(3) much when(3) it comes to thinking about reasons why not to do things {BREATH} kids can be <sil> full of inspiring aspirations and(2) hopeful thinking <sil> {NOISE} like {NOISE} my wish that no one(2) went hungry {BREATH} or that(2) everything were(2) free... | AdoraSvitak | |
i stand in these legs {COUGH} my {COUGH} hamstring and my <sil> glutes are contracted as(3) {COUGH} i {NOISE} would be had i had <sil> feet and were(2) standing on the ball of my {UM} feet <sil> it's a company in <sil> san diego called <sil> flex foot <sil> and i was(2) a(2) {NOISE} guinea pig <sil> and {UH} as i hope... | AimeeMullins | |
i looked in the {UH} rearview mirror {SMACK} and <sil> all of a sudden it just(2) hit me {UH} there was no motorcade back there {BREATH} {NOISE} you've heard of phantom(2) limb pain <sil> this was a rented(2) ford taurus {SMACK} <sil> she {SMACK} took our order and then went to the couple in the booth next to us and s... | AlGore | |
{NOISE} these are(2) real <sil> objects now <sil> i'm(2) going to show you how it is {SMACK} done {COUGH} i've looped the film here so you can(2) get a very {BREATH} interesting experience {BREATH} i want you to(2) see <sil> how this illusion is constructed(2) {UM} and it's going to rotate {BREATH} so you see that(2) ... | AlSeckel | |
{NOISE} and(2) {COUGH} uses(3) that <sil> to come to a complete vision <sil> of who you are <sil> that {SMACK} is {COUGH} snobbery {BREATH} and(2) {COUGH} the dominant {UH} kind of snobbery that(2) exists(2) nowadays {BREATH} job(2) snobbery <sil> you encounter(2) it {UM} within minutes at a party {BREATH} when(2) you... | AlaindeBotton | |
to actually(3) {NOISE} get any kind of picture <sil> on the world we live(2) in or on ourselves <sil> so what i'm(2) doing is measuring from the bottom <sil> of one image(2) to(3) the(2) <sil> bottom of {NOISE} the next image(2) about a fifth(2) of a second later <sil> <sil> like that <sil> and they're getting faster ... | AlanKay | |
and(2) that bio reactor will lay down in the wound(2) bed {BREATH} the gun that(2) you see there {NOISE} sprays cells <sil> <sil> {NOISE} that's {SMACK} going(2) to spray {UM} cells over {UH} that(2) area the reactor will serve {COUGH} to fertilize the environment {SMACK} deliver other things as(2) well at the same ti... | AlanRussell | |
i went one(2) step further and said why(2) do we have(2) to stick {SMACK} with the stodgy lawyers and just {BREATH} have(2) {UH} a <sil> paper document let's(2) go online <sil> {UM} people might need {SMACK} help in computation working with the harvard business school <sil> you'll see this example when(3) you talk abo... | AlanSiegel | |
i don't have(2) to tell you the(2) internet {SMACK} have(2) come tumbling(2) down {BREATH} and of course <sil> the(2) iron curtains have come tumbling(2) down <sil> now all of this has been tremendous {SMACK} for the world <sil> and(2) perhaps most(2) remarkably <sil> at the beginning {UH} of the twenty first century ... | AlexTabarrok | |
to seventy(2) <sil> percent {UH} at the(2) end of voting <sil> which(2) is pretty impressive right we won mister splashy pants was chosen {BREATH} hmm <sil> just kidding okay so greenpeace actually(4) wasn't that crazy about it because(2) they wanted(2) one(2) of their more thoughtful names to win so they said no no j... | AlexisOhanian |
End of preview. Expand in Data Studio
TED-LIUM Long-Form Speaker Dataset (Release 1, 100 Speakers)
This dataset is a derived version of TED-LIUM Release 1 that consolidates audio segments for 100 unique speakers into long-form WAV files. The goal is to provide a long-context speech dataset suitable for training or evaluating models on speaker-specific, continuous speech data.
🔍 Description
- Source: TED-LIUM Release 1
- Speakers: 100 (randomly selected)
- Segments: Merged per speaker across all available talks
- Split: Train
- Audio Format: 16-bit mono WAV
- Sampling Rate: 16,000 Hz
- Language: English
📁 Dataset Structure
Each example contains:
audio: Long-form audio (.wav) with all of a speaker’s utterances mergedtext: Full transcript merged from all segments of that speakerspeaker_id: Normalized speaker name (e.g.,Al_Gore)
🛠 How It Was Constructed
- Loaded TED-LIUM Release 1 using Hugging Face Datasets.
- Normalized speaker IDs by stripping suffixes (e.g.,
Al_Gore_01→Al_Gore). - Merged all audio segments and transcripts for each speaker.
- Selected 100 unique speakers (due to memory constraints).
- Saved merged
.wavfiles and transcripts into a Hugging Face-compatible dataset.
✅ Intended Use
- Long-context speech recognition
- Speaker adaptation and diarization research
- Pretraining or evaluating speech models on speaker-specific data
📜 Citation
@inproceedings{rousseau2012tedlium,
title={TED-LIUM: an Automatic Speech Recognition dedicated corpus},
author={Rousseau, Anthony and Del{\'e}glise, Paul and Est{\`e}ve, Yannick},
booktitle={Conference on Language Resources and Evaluation (LREC)},
pages={125--129},
year={2012}
}
- Downloads last month
- 9