README: EuroParl sample = 168 files, 258 h, 8 parquet shards
Browse files
README.md
CHANGED
|
@@ -4,7 +4,7 @@
|
|
| 4 |
|
| 5 |
## **Executive Summary**
|
| 6 |
|
| 7 |
-
This datacard provides a comprehensive description of **YouTube-Commons** and **EUVoxCommons** (European Parliament proceedings) collected and handled by pleias along with a sample of 1,
|
| 8 |
|
| 9 |
These datasets represent the largest collection of fully open-source copyright-compliant speech data for the 24 official languages of the European Union and more.
|
| 10 |
|
|
@@ -20,7 +20,7 @@ The combined YouTube-Commons and VoxPopuli dataset covers 23 out of 24 official
|
|
| 20 |
|
| 21 |
### **Sample description**
|
| 22 |
|
| 23 |
-
This datacard features a sample of 1,136 audio files from YouTube (394 hours) published from 2009 to 2026 in 93 languages and
|
| 24 |
|
| 25 |
Samples use the same structure as the final dataset and are distributed as fully shuffled parquet files — 12 for YouTube (`youtube_commons/`) and 8 for the European Parliament (`europarl/`) — with three components:
|
| 26 |
* Metadata scraped from the Youtube official API with original url, channel information, date of publication.
|
|
|
|
| 4 |
|
| 5 |
## **Executive Summary**
|
| 6 |
|
| 7 |
+
This datacard provides a comprehensive description of **YouTube-Commons** and **EUVoxCommons** (European Parliament proceedings) collected and handled by pleias along with a sample of 1,304 documented audio files.
|
| 8 |
|
| 9 |
These datasets represent the largest collection of fully open-source copyright-compliant speech data for the 24 official languages of the European Union and more.
|
| 10 |
|
|
|
|
| 20 |
|
| 21 |
### **Sample description**
|
| 22 |
|
| 23 |
+
This datacard features a sample of 1,136 audio files from YouTube (394 hours) published from 2009 to 2026 in 93 languages and 168 audio files from the European Parliament (258 hours): 100 VoxPopuli plenary parts from 2009–2020, and 68 recordings from 2021–2026 and from pre-2021 sessions absent from VoxPopuli (plenary sittings, committee meetings, delegations and special events), one interpretation language per recording, drawn at random. European Parliament rows carry no transcript yet.
|
| 24 |
|
| 25 |
Samples use the same structure as the final dataset and are distributed as fully shuffled parquet files — 12 for YouTube (`youtube_commons/`) and 8 for the European Parliament (`europarl/`) — with three components:
|
| 26 |
* Metadata scraped from the Youtube official API with original url, channel information, date of publication.
|