First release

First release

Yeah, we are building the largest open STT dataset for the Russian language)
Because we beilieve that life is not a zero-sum game.

This release mostly consists of our attempts to:

• See what is available;
• Gather and document work done before us in one place;

Historical composition and downloads

(Old dowload links will be discarded every iteration or two)

Type	Utterances	Hours	GB	Speaker sets	Characters	Mean length, seconds	Mean chars
Lecture	6,803	6.3	1.9	29	316,953	3.36	46.6
Narration	67,052	80.3	27.5	584	3,075,827	4.31	45.9
Phone_calls	233,868	211.2	45.9	8175	6,706,717	3.25	28.7
Series	20,243	17.5	5.2	51	759,433	3.10	37.5
Total	327,966	315	80	8,839	10,858,930

1. Dowload the chunks:

wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_aa
wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_ab
wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_ac
wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_ad
wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_ae
wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01.tar.gz_af

For multi-threaded downloads use aria2 with -x flag.

2. Download the meta data:

wget https://ru-open-stt-v01.ams3.digitaloceanspaces.com/ru_open_stt_v01_public.csv

3. Put the chunks together:
   cat ru_open_stt_v01.tar.gz_* > ru_open_stt_v01.tar.gz

4. Unpack