id stringclasses 5
values | audio audioduration (s) 6.16 12.5 | text stringclasses 5
values | voxpopuli_id stringclasses 5
values | duration_s float64 6.16 12.5 | language stringclasses 1
value | gender stringclasses 2
values | accent stringclasses 1
value | is_gold_transcript bool 1
class | voxpopuli_raw_text stringclasses 5
values | voxpopuli_normalized_text stringclasses 5
values | word_timestamps listlengths 13 19 | speaker_id stringclasses 5
values | speaker_id_source stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
vp_train_168289 | All of this is positive, yet there are very significant challenges that persist. | 20190313-0900-PLENARY-en_20190313-15:59:22_13 | 6.16 | en | female | None | true | All of this is positive, yet there are very significant challenges that persist. | all of this is positive yet there are very significant challenges that persist. | [
{
"word": "All",
"start": 0.32,
"end": 0.48,
"score": 0.9539016783237457
},
{
"word": "of",
"start": 0.48,
"end": 0.56,
"score": 0.8613452315330505
},
{
"word": "this",
"start": 0.64,
"end": 0.96,
"score": 0.8129779100418091
},
{
"word": "is",
"sta... | 28116 | voxpopuli | |
vp_train_168290 | At the same time, I want to reassure you that the EU listing has no punitive intention. | 20180205-0900-PLENARY-20-en_20180205-23:15:35_13 | 7.12 | en | female | None | true | At the same time, I want to reassure you that the EU listing has no punitive intention. | at the same time i want to reassure you that the eu listing has no punitive intention. | [
{
"word": "At",
"start": 0.32,
"end": 0.48,
"score": 0.9807382822036743
},
{
"word": "the",
"start": 0.48,
"end": 0.56,
"score": 0.9798922836780548
},
{
"word": "same",
"start": 0.56,
"end": 0.88,
"score": 0.9870938062667847
},
{
"word": "time",
"s... | inferred_00006 | inferred_cluster | |
vp_train_168291 | The Presidency expressed this sentiment through a statement which was issued shortly after President Obama signed the Executive Orders. | 20090203-0900-PLENARY-9-en_20090203-15:03:26_4 | 12.4795 | en | male | None | true | The Presidency expressed this sentiment through a statement which was issued shortly after President Obama signed the Executive Orders. | the presidency expressed this sentiment through a statement which was issued shortly after president obama signed the executive orders. | [
{
"word": "The",
"start": 0.08,
"end": 0.32,
"score": 0.4828017055988312
},
{
"word": "Presidency",
"start": 0.48,
"end": 1.2,
"score": 0.5694943368434906
},
{
"word": "expressed",
"start": 1.2,
"end": 1.84,
"score": 0.6227274239063263
},
{
"word": "th... | 95116 | voxpopuli | |
vp_train_168294 | And then state-linked media labelled these women as traitors, and they have since been charged with serious crimes. | 20180531-0900-PLENARY-4-en_20180531-10:51:33_1 | 6.3 | en | male | None | true | Then, state linked media labelled these women as traitors, and they have since been charged with serious crimes. | then state linked media labelled these women as traitors and they have since been charged with serious crimes. | [
{
"word": "And",
"start": 0.24,
"end": 0.4,
"score": 0.9512798190116882
},
{
"word": "then",
"start": 0.4,
"end": 0.64,
"score": 0.8760237395763397
},
{
"word": "statelinked",
"start": 0.8,
"end": 1.44,
"score": 0.9347124099731445
},
{
"word": "media",... | 96945 | voxpopuli | |
vp_train_168295 | Consequently, I am concerned that extending the EPO to cover civil matters would lead to confusion. | 20130522-0900-PLENARY-3-en_20130522-08:51:45_8 | 7.860063 | en | female | None | true | Consequently, I am concerned that extending the EPO to cover civil matters would lead to confusion. | consequently i am concerned that extending the epo to cover civil matters would lead to confusion. | [
{
"word": "Consequently",
"start": 0.16,
"end": 1.12,
"score": 0.9238126277923584
},
{
"word": "I",
"start": 1.36,
"end": 1.52,
"score": 0.6312651336193085
},
{
"word": "am",
"start": 1.52,
"end": 1.6,
"score": 0.8602012693881989
},
{
"word": "concerne... | 96944 | voxpopuli |
VoxPopuli-Platinum-en
Showcase page and playable sample rows for the full 142,066-row / ~404-hour English VoxPopuli Platinum dataset.
Reach out to data@trelis.com to purchase access or discuss a larger custom-curation engagement.
Training Results
These results show why the Platinum labels matter. We compare the base model,
fine-tuning on raw VoxPopuli transcripts, and fine-tuning on this internally
filtered Platinum dataset. Evaluation uses the same english-spoken corpus WER
setup across four held-out ASR test sets.
Qwen3-ASR-0.6B WER
| Training data | VP-test | FLEURS | CV-scripted | CV-spontaneous |
|---|---|---|---|---|
| Base model, no fine-tune | 2.39% | 4.40% | 7.31% | 9.16% |
| Raw VoxPopuli transcripts | 4.49% | 4.70% | 19.45% | 16.77% |
| VoxPopuli Platinum | 2.17% | 4.02% | 7.73% | 9.90% |
Whisper Large v3 Turbo WER
| Training data | VP-test | FLEURS | CV-scripted | CV-spontaneous |
|---|---|---|---|---|
| Base model, no fine-tune | 8.26% | 4.41% | 8.38% | 10.93% |
| Raw VoxPopuli transcripts | 3.77% | 7.08% | 15.75% | 24.27% |
| VoxPopuli Platinum | 3.12% | 5.90% | 11.41% | 11.36% |
Raw VoxPopuli labels are useful in-domain but can damage out-of-domain behavior. The Platinum set is filtered with Trelis internal heuristics to keep the useful VoxPopuli signal while rejecting or repairing label/audio mismatches. The training results above are the evidence for that quality filter; the filtering recipe itself is proprietary.
Contents
- 10 parquet shards:
data/train-00000.parquetthroughdata/train-00009.parquet - 142,066 rows / about 404 hours
- 16 kHz English speech from VoxPopuli
- Cleaned training text
- Word-level timestamps
- Speaker IDs for every row
- Original VoxPopuli
raw_textandnormalized_textfor auditability
Recommended Use
This dataset is suitable for ASR fine-tuning, timestamp-aware ASR experiments, captioning/alignment workflows, speaker-aware filtering, and TTS experiments. For TTS, treat it as parliamentary speech with cleaned text and speaker IDs, not as studio-recorded voice-cloning data.
Speaker IDs
| source | rows | meaning |
|---|---|---|
voxpopuli |
91,482 | Original VoxPopuli speaker ID was present and preserved |
inferred_matched |
16,104 | Missing-source row was matched to an existing speaker identity |
inferred_cluster |
34,480 | Missing-source row was assigned to an inferred speaker cluster |
The speaker_id_source column tells you whether the speaker ID came directly
from VoxPopuli or was inferred by Trelis.
Schema
| field | description |
|---|---|
id |
Trelis row ID |
audio |
16 kHz audio |
text |
cleaned ASR/TTS training text |
voxpopuli_id |
original VoxPopuli audio_id |
duration_s |
clip duration |
language |
en |
gender, accent, is_gold_transcript |
source VoxPopuli metadata |
voxpopuli_raw_text |
original VoxPopuli raw transcript text |
voxpopuli_normalized_text |
original VoxPopuli normalized transcript text |
word_timestamps |
list of {word, start, end, score} |
speaker_id |
preserved or inferred speaker ID |
speaker_id_source |
voxpopuli, inferred_matched, or inferred_cluster |
License
Commercial Trelis license required. Redistribution is not permitted.
- Downloads last month
- 98